| { | |
| "best_metric": 0.4207764409383577, | |
| "best_model_checkpoint": "/storage/elsayedissa/whisper-large-v2-arabic-24h/checkpoint-3000", | |
| "epoch": 8.9126559714795, | |
| "global_step": 5000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 4.4e-07, | |
| "loss": 0.8858, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.200000000000001e-07, | |
| "loss": 0.7395, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.4000000000000001e-06, | |
| "loss": 0.3384, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.9000000000000002e-06, | |
| "loss": 0.3633, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 2.4000000000000003e-06, | |
| "loss": 0.2796, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 2.9e-06, | |
| "loss": 0.3068, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 3.4000000000000005e-06, | |
| "loss": 0.251, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 3.900000000000001e-06, | |
| "loss": 0.2845, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 4.4e-06, | |
| "loss": 0.244, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 4.9000000000000005e-06, | |
| "loss": 0.297, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 5.400000000000001e-06, | |
| "loss": 0.2282, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 5.9e-06, | |
| "loss": 0.2681, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 6.4000000000000006e-06, | |
| "loss": 0.2311, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.9e-06, | |
| "loss": 0.265, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 7.4e-06, | |
| "loss": 0.2299, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 7.9e-06, | |
| "loss": 0.2707, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 8.400000000000001e-06, | |
| "loss": 0.2071, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 8.900000000000001e-06, | |
| "loss": 0.2662, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 9.4e-06, | |
| "loss": 0.2056, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 9.9e-06, | |
| "loss": 0.2588, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 9.955555555555556e-06, | |
| "loss": 0.2123, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 9.9e-06, | |
| "loss": 0.247, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 9.844444444444446e-06, | |
| "loss": 0.1877, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 9.78888888888889e-06, | |
| "loss": 0.1774, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 9.733333333333334e-06, | |
| "loss": 0.1732, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 9.677777777777778e-06, | |
| "loss": 0.1739, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 9.622222222222222e-06, | |
| "loss": 0.1613, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 9.566666666666668e-06, | |
| "loss": 0.1711, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 9.511111111111112e-06, | |
| "loss": 0.1604, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 9.455555555555557e-06, | |
| "loss": 0.1623, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 9.4e-06, | |
| "loss": 0.1584, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 9.344444444444446e-06, | |
| "loss": 0.1664, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 9.28888888888889e-06, | |
| "loss": 0.1708, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 9.233333333333334e-06, | |
| "loss": 0.1647, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 9.17777777777778e-06, | |
| "loss": 0.1595, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 9.122222222222223e-06, | |
| "loss": 0.1599, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "learning_rate": 9.066666666666667e-06, | |
| "loss": 0.1686, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 9.011111111111111e-06, | |
| "loss": 0.1684, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 8.955555555555555e-06, | |
| "loss": 0.1566, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 8.900000000000001e-06, | |
| "loss": 0.1638, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "eval_loss": 0.22948628664016724, | |
| "eval_runtime": 4406.6894, | |
| "eval_samples_per_second": 2.369, | |
| "eval_steps_per_second": 0.148, | |
| "eval_wer": 0.4409728815314771, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 8.844444444444445e-06, | |
| "loss": 0.1597, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 8.788888888888891e-06, | |
| "loss": 0.1524, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 8.733333333333333e-06, | |
| "loss": 0.1527, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 8.677777777777779e-06, | |
| "loss": 0.151, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 8.622222222222223e-06, | |
| "loss": 0.158, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 8.566666666666667e-06, | |
| "loss": 0.0797, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 8.511111111111113e-06, | |
| "loss": 0.1002, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 8.455555555555555e-06, | |
| "loss": 0.096, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 8.400000000000001e-06, | |
| "loss": 0.098, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 8.344444444444445e-06, | |
| "loss": 0.0927, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 8.288888888888889e-06, | |
| "loss": 0.1003, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 2.32, | |
| "learning_rate": 8.233333333333335e-06, | |
| "loss": 0.0885, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 8.177777777777779e-06, | |
| "loss": 0.1033, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 2.41, | |
| "learning_rate": 8.122222222222223e-06, | |
| "loss": 0.0914, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 2.45, | |
| "learning_rate": 8.066666666666667e-06, | |
| "loss": 0.1049, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 8.011111111111113e-06, | |
| "loss": 0.0889, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 2.54, | |
| "learning_rate": 7.955555555555557e-06, | |
| "loss": 0.0989, | |
| "step": 1425 | |
| }, | |
| { | |
| "epoch": 2.58, | |
| "learning_rate": 7.9e-06, | |
| "loss": 0.0952, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 2.63, | |
| "learning_rate": 7.844444444444446e-06, | |
| "loss": 0.1034, | |
| "step": 1475 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 7.788888888888889e-06, | |
| "loss": 0.0906, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 2.72, | |
| "learning_rate": 7.733333333333334e-06, | |
| "loss": 0.0946, | |
| "step": 1525 | |
| }, | |
| { | |
| "epoch": 2.76, | |
| "learning_rate": 7.677777777777778e-06, | |
| "loss": 0.0865, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 2.81, | |
| "learning_rate": 7.622222222222223e-06, | |
| "loss": 0.105, | |
| "step": 1575 | |
| }, | |
| { | |
| "epoch": 2.85, | |
| "learning_rate": 7.566666666666667e-06, | |
| "loss": 0.0891, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 2.9, | |
| "learning_rate": 7.511111111111111e-06, | |
| "loss": 0.1019, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 2.94, | |
| "learning_rate": 7.455555555555556e-06, | |
| "loss": 0.09, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "learning_rate": 7.4e-06, | |
| "loss": 0.0941, | |
| "step": 1675 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 7.344444444444445e-06, | |
| "loss": 0.0618, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "learning_rate": 7.28888888888889e-06, | |
| "loss": 0.0489, | |
| "step": 1725 | |
| }, | |
| { | |
| "epoch": 3.12, | |
| "learning_rate": 7.233333333333334e-06, | |
| "loss": 0.0523, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 3.16, | |
| "learning_rate": 7.177777777777778e-06, | |
| "loss": 0.056, | |
| "step": 1775 | |
| }, | |
| { | |
| "epoch": 3.21, | |
| "learning_rate": 7.122222222222222e-06, | |
| "loss": 0.0544, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 3.25, | |
| "learning_rate": 7.066666666666667e-06, | |
| "loss": 0.0535, | |
| "step": 1825 | |
| }, | |
| { | |
| "epoch": 3.3, | |
| "learning_rate": 7.011111111111112e-06, | |
| "loss": 0.0548, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 3.34, | |
| "learning_rate": 6.955555555555557e-06, | |
| "loss": 0.0566, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 3.39, | |
| "learning_rate": 6.9e-06, | |
| "loss": 0.0563, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 3.43, | |
| "learning_rate": 6.844444444444445e-06, | |
| "loss": 0.0566, | |
| "step": 1925 | |
| }, | |
| { | |
| "epoch": 3.48, | |
| "learning_rate": 6.788888888888889e-06, | |
| "loss": 0.0588, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 3.52, | |
| "learning_rate": 6.733333333333334e-06, | |
| "loss": 0.0555, | |
| "step": 1975 | |
| }, | |
| { | |
| "epoch": 3.57, | |
| "learning_rate": 6.677777777777779e-06, | |
| "loss": 0.0587, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 3.57, | |
| "eval_loss": 0.23366238176822662, | |
| "eval_runtime": 4445.0678, | |
| "eval_samples_per_second": 2.348, | |
| "eval_steps_per_second": 0.147, | |
| "eval_wer": 0.42719787333206743, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 3.61, | |
| "learning_rate": 6.6222222222222236e-06, | |
| "loss": 0.051, | |
| "step": 2025 | |
| }, | |
| { | |
| "epoch": 3.65, | |
| "learning_rate": 6.566666666666667e-06, | |
| "loss": 0.0552, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 3.7, | |
| "learning_rate": 6.511111111111112e-06, | |
| "loss": 0.0559, | |
| "step": 2075 | |
| }, | |
| { | |
| "epoch": 3.74, | |
| "learning_rate": 6.455555555555556e-06, | |
| "loss": 0.0562, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 3.79, | |
| "learning_rate": 6.4000000000000006e-06, | |
| "loss": 0.0597, | |
| "step": 2125 | |
| }, | |
| { | |
| "epoch": 3.83, | |
| "learning_rate": 6.3444444444444454e-06, | |
| "loss": 0.0512, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 3.88, | |
| "learning_rate": 6.28888888888889e-06, | |
| "loss": 0.0533, | |
| "step": 2175 | |
| }, | |
| { | |
| "epoch": 3.92, | |
| "learning_rate": 6.2333333333333335e-06, | |
| "loss": 0.0545, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 3.97, | |
| "learning_rate": 6.177777777777778e-06, | |
| "loss": 0.0538, | |
| "step": 2225 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "learning_rate": 6.124444444444445e-06, | |
| "loss": 0.05, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 4.06, | |
| "learning_rate": 6.06888888888889e-06, | |
| "loss": 0.0303, | |
| "step": 2275 | |
| }, | |
| { | |
| "epoch": 4.1, | |
| "learning_rate": 6.013333333333335e-06, | |
| "loss": 0.0259, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 4.14, | |
| "learning_rate": 5.957777777777778e-06, | |
| "loss": 0.0257, | |
| "step": 2325 | |
| }, | |
| { | |
| "epoch": 4.19, | |
| "learning_rate": 5.902222222222223e-06, | |
| "loss": 0.0303, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 4.23, | |
| "learning_rate": 5.846666666666667e-06, | |
| "loss": 0.0292, | |
| "step": 2375 | |
| }, | |
| { | |
| "epoch": 4.28, | |
| "learning_rate": 5.791111111111112e-06, | |
| "loss": 0.0307, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 4.32, | |
| "learning_rate": 5.735555555555557e-06, | |
| "loss": 0.0274, | |
| "step": 2425 | |
| }, | |
| { | |
| "epoch": 4.37, | |
| "learning_rate": 5.68e-06, | |
| "loss": 0.0277, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 4.41, | |
| "learning_rate": 5.624444444444445e-06, | |
| "loss": 0.0277, | |
| "step": 2475 | |
| }, | |
| { | |
| "epoch": 4.46, | |
| "learning_rate": 5.56888888888889e-06, | |
| "loss": 0.0327, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 4.5, | |
| "learning_rate": 5.513333333333334e-06, | |
| "loss": 0.0268, | |
| "step": 2525 | |
| }, | |
| { | |
| "epoch": 4.55, | |
| "learning_rate": 5.4577777777777785e-06, | |
| "loss": 0.0317, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 4.59, | |
| "learning_rate": 5.402222222222223e-06, | |
| "loss": 0.0289, | |
| "step": 2575 | |
| }, | |
| { | |
| "epoch": 4.63, | |
| "learning_rate": 5.346666666666667e-06, | |
| "loss": 0.0304, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 4.68, | |
| "learning_rate": 5.2911111111111115e-06, | |
| "loss": 0.0304, | |
| "step": 2625 | |
| }, | |
| { | |
| "epoch": 4.72, | |
| "learning_rate": 5.235555555555556e-06, | |
| "loss": 0.0286, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 4.77, | |
| "learning_rate": 5.18e-06, | |
| "loss": 0.0285, | |
| "step": 2675 | |
| }, | |
| { | |
| "epoch": 4.81, | |
| "learning_rate": 5.124444444444445e-06, | |
| "loss": 0.0303, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 4.86, | |
| "learning_rate": 5.06888888888889e-06, | |
| "loss": 0.0281, | |
| "step": 2725 | |
| }, | |
| { | |
| "epoch": 4.9, | |
| "learning_rate": 5.013333333333333e-06, | |
| "loss": 0.0291, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 4.95, | |
| "learning_rate": 4.957777777777778e-06, | |
| "loss": 0.0286, | |
| "step": 2775 | |
| }, | |
| { | |
| "epoch": 4.99, | |
| "learning_rate": 4.902222222222222e-06, | |
| "loss": 0.0298, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 5.04, | |
| "learning_rate": 4.846666666666667e-06, | |
| "loss": 0.0238, | |
| "step": 2825 | |
| }, | |
| { | |
| "epoch": 5.08, | |
| "learning_rate": 4.791111111111111e-06, | |
| "loss": 0.0122, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 5.12, | |
| "learning_rate": 4.735555555555556e-06, | |
| "loss": 0.0148, | |
| "step": 2875 | |
| }, | |
| { | |
| "epoch": 5.17, | |
| "learning_rate": 4.680000000000001e-06, | |
| "loss": 0.0123, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 5.21, | |
| "learning_rate": 4.624444444444445e-06, | |
| "loss": 0.0162, | |
| "step": 2925 | |
| }, | |
| { | |
| "epoch": 5.26, | |
| "learning_rate": 4.568888888888889e-06, | |
| "loss": 0.015, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 5.3, | |
| "learning_rate": 4.513333333333333e-06, | |
| "loss": 0.0161, | |
| "step": 2975 | |
| }, | |
| { | |
| "epoch": 5.35, | |
| "learning_rate": 4.457777777777778e-06, | |
| "loss": 0.0125, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 5.35, | |
| "eval_loss": 0.2745016813278198, | |
| "eval_runtime": 4396.7332, | |
| "eval_samples_per_second": 2.374, | |
| "eval_steps_per_second": 0.149, | |
| "eval_wer": 0.4207764409383577, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 5.39, | |
| "learning_rate": 4.402222222222223e-06, | |
| "loss": 0.0152, | |
| "step": 3025 | |
| }, | |
| { | |
| "epoch": 5.44, | |
| "learning_rate": 4.346666666666667e-06, | |
| "loss": 0.0143, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 5.48, | |
| "learning_rate": 4.291111111111112e-06, | |
| "loss": 0.0143, | |
| "step": 3075 | |
| }, | |
| { | |
| "epoch": 5.53, | |
| "learning_rate": 4.235555555555556e-06, | |
| "loss": 0.0136, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 5.57, | |
| "learning_rate": 4.18e-06, | |
| "loss": 0.0173, | |
| "step": 3125 | |
| }, | |
| { | |
| "epoch": 5.61, | |
| "learning_rate": 4.124444444444445e-06, | |
| "loss": 0.0136, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 5.66, | |
| "learning_rate": 4.0688888888888896e-06, | |
| "loss": 0.0154, | |
| "step": 3175 | |
| }, | |
| { | |
| "epoch": 5.7, | |
| "learning_rate": 4.013333333333334e-06, | |
| "loss": 0.0119, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 5.75, | |
| "learning_rate": 3.9577777777777785e-06, | |
| "loss": 0.0138, | |
| "step": 3225 | |
| }, | |
| { | |
| "epoch": 5.79, | |
| "learning_rate": 3.9022222222222225e-06, | |
| "loss": 0.0131, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 5.84, | |
| "learning_rate": 3.8466666666666665e-06, | |
| "loss": 0.0154, | |
| "step": 3275 | |
| }, | |
| { | |
| "epoch": 5.88, | |
| "learning_rate": 3.7911111111111114e-06, | |
| "loss": 0.0134, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 5.93, | |
| "learning_rate": 3.7355555555555555e-06, | |
| "loss": 0.0145, | |
| "step": 3325 | |
| }, | |
| { | |
| "epoch": 5.97, | |
| "learning_rate": 3.6800000000000003e-06, | |
| "loss": 0.0125, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 6.02, | |
| "learning_rate": 3.624444444444445e-06, | |
| "loss": 0.0111, | |
| "step": 3375 | |
| }, | |
| { | |
| "epoch": 6.06, | |
| "learning_rate": 3.568888888888889e-06, | |
| "loss": 0.0072, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 6.11, | |
| "learning_rate": 3.5133333333333337e-06, | |
| "loss": 0.0072, | |
| "step": 3425 | |
| }, | |
| { | |
| "epoch": 6.15, | |
| "learning_rate": 3.457777777777778e-06, | |
| "loss": 0.0068, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 6.19, | |
| "learning_rate": 3.4022222222222222e-06, | |
| "loss": 0.0076, | |
| "step": 3475 | |
| }, | |
| { | |
| "epoch": 6.24, | |
| "learning_rate": 3.346666666666667e-06, | |
| "loss": 0.0068, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 6.28, | |
| "learning_rate": 3.2911111111111116e-06, | |
| "loss": 0.0077, | |
| "step": 3525 | |
| }, | |
| { | |
| "epoch": 6.33, | |
| "learning_rate": 3.2355555555555556e-06, | |
| "loss": 0.0072, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 6.37, | |
| "learning_rate": 3.1800000000000005e-06, | |
| "loss": 0.0082, | |
| "step": 3575 | |
| }, | |
| { | |
| "epoch": 6.42, | |
| "learning_rate": 3.124444444444445e-06, | |
| "loss": 0.0064, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 6.46, | |
| "learning_rate": 3.068888888888889e-06, | |
| "loss": 0.0072, | |
| "step": 3625 | |
| }, | |
| { | |
| "epoch": 6.51, | |
| "learning_rate": 3.013333333333334e-06, | |
| "loss": 0.0063, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 6.55, | |
| "learning_rate": 2.957777777777778e-06, | |
| "loss": 0.0082, | |
| "step": 3675 | |
| }, | |
| { | |
| "epoch": 6.6, | |
| "learning_rate": 2.9022222222222223e-06, | |
| "loss": 0.0054, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 6.64, | |
| "learning_rate": 2.8466666666666672e-06, | |
| "loss": 0.0074, | |
| "step": 3725 | |
| }, | |
| { | |
| "epoch": 6.68, | |
| "learning_rate": 2.7911111111111113e-06, | |
| "loss": 0.006, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 6.73, | |
| "learning_rate": 2.7355555555555557e-06, | |
| "loss": 0.0074, | |
| "step": 3775 | |
| }, | |
| { | |
| "epoch": 6.77, | |
| "learning_rate": 2.68e-06, | |
| "loss": 0.0061, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 6.82, | |
| "learning_rate": 2.6244444444444446e-06, | |
| "loss": 0.0089, | |
| "step": 3825 | |
| }, | |
| { | |
| "epoch": 6.86, | |
| "learning_rate": 2.568888888888889e-06, | |
| "loss": 0.0066, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 6.91, | |
| "learning_rate": 2.5133333333333336e-06, | |
| "loss": 0.0071, | |
| "step": 3875 | |
| }, | |
| { | |
| "epoch": 6.95, | |
| "learning_rate": 2.457777777777778e-06, | |
| "loss": 0.0065, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "learning_rate": 2.4022222222222225e-06, | |
| "loss": 0.0053, | |
| "step": 3925 | |
| }, | |
| { | |
| "epoch": 7.04, | |
| "learning_rate": 2.346666666666667e-06, | |
| "loss": 0.0049, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 7.09, | |
| "learning_rate": 2.2911111111111114e-06, | |
| "loss": 0.0021, | |
| "step": 3975 | |
| }, | |
| { | |
| "epoch": 7.13, | |
| "learning_rate": 2.235555555555556e-06, | |
| "loss": 0.004, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 7.13, | |
| "eval_loss": 0.3123507499694824, | |
| "eval_runtime": 4382.7686, | |
| "eval_samples_per_second": 2.382, | |
| "eval_steps_per_second": 0.149, | |
| "eval_wer": 0.4251609673577187, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 7.17, | |
| "learning_rate": 2.1800000000000003e-06, | |
| "loss": 0.002, | |
| "step": 4025 | |
| }, | |
| { | |
| "epoch": 7.22, | |
| "learning_rate": 2.1244444444444443e-06, | |
| "loss": 0.0042, | |
| "step": 4050 | |
| }, | |
| { | |
| "epoch": 7.26, | |
| "learning_rate": 2.0688888888888892e-06, | |
| "loss": 0.0021, | |
| "step": 4075 | |
| }, | |
| { | |
| "epoch": 7.31, | |
| "learning_rate": 2.0133333333333337e-06, | |
| "loss": 0.0035, | |
| "step": 4100 | |
| }, | |
| { | |
| "epoch": 7.35, | |
| "learning_rate": 1.9577777777777777e-06, | |
| "loss": 0.0026, | |
| "step": 4125 | |
| }, | |
| { | |
| "epoch": 7.4, | |
| "learning_rate": 1.9022222222222222e-06, | |
| "loss": 0.0037, | |
| "step": 4150 | |
| }, | |
| { | |
| "epoch": 7.44, | |
| "learning_rate": 1.8466666666666668e-06, | |
| "loss": 0.0024, | |
| "step": 4175 | |
| }, | |
| { | |
| "epoch": 7.49, | |
| "learning_rate": 1.7911111111111113e-06, | |
| "loss": 0.0038, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 7.53, | |
| "learning_rate": 1.7355555555555555e-06, | |
| "loss": 0.0024, | |
| "step": 4225 | |
| }, | |
| { | |
| "epoch": 7.58, | |
| "learning_rate": 1.6800000000000002e-06, | |
| "loss": 0.0036, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 7.62, | |
| "learning_rate": 1.6244444444444447e-06, | |
| "loss": 0.0028, | |
| "step": 4275 | |
| }, | |
| { | |
| "epoch": 7.66, | |
| "learning_rate": 1.568888888888889e-06, | |
| "loss": 0.0043, | |
| "step": 4300 | |
| }, | |
| { | |
| "epoch": 7.71, | |
| "learning_rate": 1.5133333333333334e-06, | |
| "loss": 0.0017, | |
| "step": 4325 | |
| }, | |
| { | |
| "epoch": 7.75, | |
| "learning_rate": 1.457777777777778e-06, | |
| "loss": 0.0035, | |
| "step": 4350 | |
| }, | |
| { | |
| "epoch": 7.8, | |
| "learning_rate": 1.4022222222222223e-06, | |
| "loss": 0.003, | |
| "step": 4375 | |
| }, | |
| { | |
| "epoch": 7.84, | |
| "learning_rate": 1.3466666666666668e-06, | |
| "loss": 0.0037, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 7.89, | |
| "learning_rate": 1.2911111111111112e-06, | |
| "loss": 0.0018, | |
| "step": 4425 | |
| }, | |
| { | |
| "epoch": 7.93, | |
| "learning_rate": 1.2355555555555557e-06, | |
| "loss": 0.004, | |
| "step": 4450 | |
| }, | |
| { | |
| "epoch": 7.98, | |
| "learning_rate": 1.1800000000000001e-06, | |
| "loss": 0.002, | |
| "step": 4475 | |
| }, | |
| { | |
| "epoch": 8.02, | |
| "learning_rate": 1.1244444444444446e-06, | |
| "loss": 0.0025, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 8.07, | |
| "learning_rate": 1.068888888888889e-06, | |
| "loss": 0.0014, | |
| "step": 4525 | |
| }, | |
| { | |
| "epoch": 8.11, | |
| "learning_rate": 1.0133333333333333e-06, | |
| "loss": 0.0018, | |
| "step": 4550 | |
| }, | |
| { | |
| "epoch": 8.16, | |
| "learning_rate": 9.57777777777778e-07, | |
| "loss": 0.0012, | |
| "step": 4575 | |
| }, | |
| { | |
| "epoch": 8.2, | |
| "learning_rate": 9.022222222222222e-07, | |
| "loss": 0.0022, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 8.24, | |
| "learning_rate": 8.466666666666668e-07, | |
| "loss": 0.0011, | |
| "step": 4625 | |
| }, | |
| { | |
| "epoch": 8.29, | |
| "learning_rate": 7.911111111111111e-07, | |
| "loss": 0.0018, | |
| "step": 4650 | |
| }, | |
| { | |
| "epoch": 8.33, | |
| "learning_rate": 7.355555555555556e-07, | |
| "loss": 0.0013, | |
| "step": 4675 | |
| }, | |
| { | |
| "epoch": 8.38, | |
| "learning_rate": 6.800000000000001e-07, | |
| "loss": 0.0016, | |
| "step": 4700 | |
| }, | |
| { | |
| "epoch": 8.42, | |
| "learning_rate": 6.244444444444445e-07, | |
| "loss": 0.0013, | |
| "step": 4725 | |
| }, | |
| { | |
| "epoch": 8.47, | |
| "learning_rate": 5.68888888888889e-07, | |
| "loss": 0.0019, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 8.51, | |
| "learning_rate": 5.133333333333334e-07, | |
| "loss": 0.0013, | |
| "step": 4775 | |
| }, | |
| { | |
| "epoch": 8.56, | |
| "learning_rate": 4.5777777777777784e-07, | |
| "loss": 0.0017, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 8.6, | |
| "learning_rate": 4.0222222222222224e-07, | |
| "loss": 0.0013, | |
| "step": 4825 | |
| }, | |
| { | |
| "epoch": 8.65, | |
| "learning_rate": 3.466666666666667e-07, | |
| "loss": 0.0021, | |
| "step": 4850 | |
| }, | |
| { | |
| "epoch": 8.69, | |
| "learning_rate": 2.9111111111111116e-07, | |
| "loss": 0.0012, | |
| "step": 4875 | |
| }, | |
| { | |
| "epoch": 8.73, | |
| "learning_rate": 2.3555555555555556e-07, | |
| "loss": 0.0026, | |
| "step": 4900 | |
| }, | |
| { | |
| "epoch": 8.78, | |
| "learning_rate": 1.8e-07, | |
| "loss": 0.0012, | |
| "step": 4925 | |
| }, | |
| { | |
| "epoch": 8.82, | |
| "learning_rate": 1.2444444444444446e-07, | |
| "loss": 0.0017, | |
| "step": 4950 | |
| }, | |
| { | |
| "epoch": 8.87, | |
| "learning_rate": 6.888888888888889e-08, | |
| "loss": 0.0012, | |
| "step": 4975 | |
| }, | |
| { | |
| "epoch": 8.91, | |
| "learning_rate": 1.3333333333333334e-08, | |
| "loss": 0.0016, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 8.91, | |
| "eval_loss": 0.3434337377548218, | |
| "eval_runtime": 4386.6592, | |
| "eval_samples_per_second": 2.379, | |
| "eval_steps_per_second": 0.149, | |
| "eval_wer": 0.4239008475600283, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 8.91, | |
| "step": 5000, | |
| "total_flos": 5.306738576228352e+20, | |
| "train_loss": 0.07621470112651586, | |
| "train_runtime": 92554.3874, | |
| "train_samples_per_second": 2.701, | |
| "train_steps_per_second": 0.054 | |
| } | |
| ], | |
| "max_steps": 5000, | |
| "num_train_epochs": 9, | |
| "total_flos": 5.306738576228352e+20, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |