| { | |
| "best_metric": 70.0375841272616, | |
| "best_model_checkpoint": "/exp/whisper_yue/finetune-whisper-canto/cn_model_out_trpro/checkpoint-3000", | |
| "epoch": 8.077481840193705, | |
| "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": 4.6000000000000004e-07, | |
| "loss": 2.9087, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.600000000000001e-07, | |
| "loss": 2.0508, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.46e-06, | |
| "loss": 1.4301, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9600000000000003e-06, | |
| "loss": 0.7824, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 2.46e-06, | |
| "loss": 0.5767, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 2.96e-06, | |
| "loss": 0.5376, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 3.46e-06, | |
| "loss": 0.5437, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 3.96e-06, | |
| "loss": 0.5106, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 4.4600000000000005e-06, | |
| "loss": 0.5149, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 4.960000000000001e-06, | |
| "loss": 0.4984, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 5.460000000000001e-06, | |
| "loss": 0.4944, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 5.9600000000000005e-06, | |
| "loss": 0.4829, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 6.460000000000001e-06, | |
| "loss": 0.5186, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 6.96e-06, | |
| "loss": 0.4607, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.4600000000000006e-06, | |
| "loss": 0.4705, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 7.960000000000002e-06, | |
| "loss": 0.4611, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 8.46e-06, | |
| "loss": 0.4871, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 8.96e-06, | |
| "loss": 0.4688, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 9.460000000000001e-06, | |
| "loss": 0.4874, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 9.960000000000001e-06, | |
| "loss": 0.4447, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 9.94888888888889e-06, | |
| "loss": 0.4526, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 9.893333333333334e-06, | |
| "loss": 0.4407, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 9.837777777777778e-06, | |
| "loss": 0.4499, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 9.782222222222222e-06, | |
| "loss": 0.4415, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 9.726666666666668e-06, | |
| "loss": 0.463, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 9.671111111111112e-06, | |
| "loss": 0.3825, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 9.615555555555558e-06, | |
| "loss": 0.3622, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 9.56e-06, | |
| "loss": 0.3591, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 9.504444444444446e-06, | |
| "loss": 0.3754, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 9.44888888888889e-06, | |
| "loss": 0.3325, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 9.393333333333334e-06, | |
| "loss": 0.3685, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 9.33777777777778e-06, | |
| "loss": 0.3533, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 9.282222222222222e-06, | |
| "loss": 0.348, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 9.226666666666668e-06, | |
| "loss": 0.3708, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 9.171111111111112e-06, | |
| "loss": 0.3254, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 9.115555555555556e-06, | |
| "loss": 0.3627, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 9.060000000000001e-06, | |
| "loss": 0.3939, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 9.004444444444445e-06, | |
| "loss": 0.3532, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 8.94888888888889e-06, | |
| "loss": 0.3254, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 8.893333333333333e-06, | |
| "loss": 0.3296, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "eval_loss": 0.33027762174606323, | |
| "eval_runtime": 2666.6434, | |
| "eval_samples_per_second": 3.968, | |
| "eval_steps_per_second": 0.248, | |
| "eval_wer": 72.6160300673018, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 8.83777777777778e-06, | |
| "loss": 0.3648, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 8.782222222222223e-06, | |
| "loss": 0.3344, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 8.726666666666667e-06, | |
| "loss": 0.3378, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 8.671111111111113e-06, | |
| "loss": 0.3489, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 8.615555555555555e-06, | |
| "loss": 0.3847, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 8.560000000000001e-06, | |
| "loss": 0.3437, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 8.504444444444445e-06, | |
| "loss": 0.3484, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 8.448888888888889e-06, | |
| "loss": 0.3471, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 8.393333333333335e-06, | |
| "loss": 0.3365, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 8.337777777777777e-06, | |
| "loss": 0.3223, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 8.282222222222223e-06, | |
| "loss": 0.256, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 8.226666666666667e-06, | |
| "loss": 0.248, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 8.171111111111113e-06, | |
| "loss": 0.254, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 8.115555555555557e-06, | |
| "loss": 0.2681, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 8.06e-06, | |
| "loss": 0.2564, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 8.004444444444445e-06, | |
| "loss": 0.267, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 7.948888888888889e-06, | |
| "loss": 0.2604, | |
| "step": 1425 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 7.893333333333335e-06, | |
| "loss": 0.2406, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 2.38, | |
| "learning_rate": 7.837777777777779e-06, | |
| "loss": 0.2564, | |
| "step": 1475 | |
| }, | |
| { | |
| "epoch": 2.42, | |
| "learning_rate": 7.782222222222223e-06, | |
| "loss": 0.2359, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 7.726666666666667e-06, | |
| "loss": 0.2549, | |
| "step": 1525 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 7.67111111111111e-06, | |
| "loss": 0.2507, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 2.54, | |
| "learning_rate": 7.6155555555555564e-06, | |
| "loss": 0.248, | |
| "step": 1575 | |
| }, | |
| { | |
| "epoch": 2.58, | |
| "learning_rate": 7.5600000000000005e-06, | |
| "loss": 0.2726, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 2.62, | |
| "learning_rate": 7.504444444444445e-06, | |
| "loss": 0.2398, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 7.44888888888889e-06, | |
| "loss": 0.2508, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 2.71, | |
| "learning_rate": 7.393333333333333e-06, | |
| "loss": 0.2638, | |
| "step": 1675 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 7.337777777777778e-06, | |
| "loss": 0.2686, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 2.79, | |
| "learning_rate": 7.282222222222222e-06, | |
| "loss": 0.2449, | |
| "step": 1725 | |
| }, | |
| { | |
| "epoch": 2.83, | |
| "learning_rate": 7.226666666666667e-06, | |
| "loss": 0.2651, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 2.87, | |
| "learning_rate": 7.171111111111112e-06, | |
| "loss": 0.2311, | |
| "step": 1775 | |
| }, | |
| { | |
| "epoch": 2.91, | |
| "learning_rate": 7.115555555555557e-06, | |
| "loss": 0.2408, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 2.95, | |
| "learning_rate": 7.06e-06, | |
| "loss": 0.2595, | |
| "step": 1825 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "learning_rate": 7.004444444444445e-06, | |
| "loss": 0.2391, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 6.948888888888889e-06, | |
| "loss": 0.2098, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "learning_rate": 6.893333333333334e-06, | |
| "loss": 0.1609, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "learning_rate": 6.837777777777779e-06, | |
| "loss": 0.1742, | |
| "step": 1925 | |
| }, | |
| { | |
| "epoch": 3.15, | |
| "learning_rate": 6.782222222222222e-06, | |
| "loss": 0.1875, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 3.19, | |
| "learning_rate": 6.726666666666667e-06, | |
| "loss": 0.1736, | |
| "step": 1975 | |
| }, | |
| { | |
| "epoch": 3.23, | |
| "learning_rate": 6.671111111111112e-06, | |
| "loss": 0.197, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 3.23, | |
| "eval_loss": 0.3263445198535919, | |
| "eval_runtime": 2726.4106, | |
| "eval_samples_per_second": 3.881, | |
| "eval_steps_per_second": 0.243, | |
| "eval_wer": 70.59697578882965, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 3.27, | |
| "learning_rate": 6.615555555555556e-06, | |
| "loss": 0.2145, | |
| "step": 2025 | |
| }, | |
| { | |
| "epoch": 3.31, | |
| "learning_rate": 6.560000000000001e-06, | |
| "loss": 0.1779, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 3.35, | |
| "learning_rate": 6.504444444444446e-06, | |
| "loss": 0.1989, | |
| "step": 2075 | |
| }, | |
| { | |
| "epoch": 3.39, | |
| "learning_rate": 6.448888888888889e-06, | |
| "loss": 0.1829, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 3.43, | |
| "learning_rate": 6.393333333333334e-06, | |
| "loss": 0.1699, | |
| "step": 2125 | |
| }, | |
| { | |
| "epoch": 3.47, | |
| "learning_rate": 6.3377777777777786e-06, | |
| "loss": 0.1858, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 3.51, | |
| "learning_rate": 6.282222222222223e-06, | |
| "loss": 0.1769, | |
| "step": 2175 | |
| }, | |
| { | |
| "epoch": 3.55, | |
| "learning_rate": 6.2266666666666675e-06, | |
| "loss": 0.1731, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 3.59, | |
| "learning_rate": 6.171111111111112e-06, | |
| "loss": 0.2021, | |
| "step": 2225 | |
| }, | |
| { | |
| "epoch": 3.63, | |
| "learning_rate": 6.1155555555555555e-06, | |
| "loss": 0.163, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 3.67, | |
| "learning_rate": 6.0600000000000004e-06, | |
| "loss": 0.1698, | |
| "step": 2275 | |
| }, | |
| { | |
| "epoch": 3.72, | |
| "learning_rate": 6.004444444444445e-06, | |
| "loss": 0.2006, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 3.76, | |
| "learning_rate": 5.948888888888889e-06, | |
| "loss": 0.1812, | |
| "step": 2325 | |
| }, | |
| { | |
| "epoch": 3.8, | |
| "learning_rate": 5.893333333333334e-06, | |
| "loss": 0.1819, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 3.84, | |
| "learning_rate": 5.837777777777777e-06, | |
| "loss": 0.1626, | |
| "step": 2375 | |
| }, | |
| { | |
| "epoch": 3.88, | |
| "learning_rate": 5.782222222222222e-06, | |
| "loss": 0.1733, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 3.92, | |
| "learning_rate": 5.726666666666667e-06, | |
| "loss": 0.2106, | |
| "step": 2425 | |
| }, | |
| { | |
| "epoch": 3.96, | |
| "learning_rate": 5.671111111111112e-06, | |
| "loss": 0.1793, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 5.615555555555556e-06, | |
| "loss": 0.175, | |
| "step": 2475 | |
| }, | |
| { | |
| "epoch": 4.04, | |
| "learning_rate": 5.560000000000001e-06, | |
| "loss": 0.1231, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 4.08, | |
| "learning_rate": 5.504444444444444e-06, | |
| "loss": 0.14, | |
| "step": 2525 | |
| }, | |
| { | |
| "epoch": 4.12, | |
| "learning_rate": 5.448888888888889e-06, | |
| "loss": 0.1273, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 4.16, | |
| "learning_rate": 5.393333333333334e-06, | |
| "loss": 0.1228, | |
| "step": 2575 | |
| }, | |
| { | |
| "epoch": 4.2, | |
| "learning_rate": 5.337777777777779e-06, | |
| "loss": 0.1246, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 4.24, | |
| "learning_rate": 5.282222222222223e-06, | |
| "loss": 0.1386, | |
| "step": 2625 | |
| }, | |
| { | |
| "epoch": 4.28, | |
| "learning_rate": 5.226666666666667e-06, | |
| "loss": 0.1306, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 4.32, | |
| "learning_rate": 5.171111111111111e-06, | |
| "loss": 0.1343, | |
| "step": 2675 | |
| }, | |
| { | |
| "epoch": 4.36, | |
| "learning_rate": 5.115555555555556e-06, | |
| "loss": 0.157, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 4.4, | |
| "learning_rate": 5.060000000000001e-06, | |
| "loss": 0.1233, | |
| "step": 2725 | |
| }, | |
| { | |
| "epoch": 4.44, | |
| "learning_rate": 5.004444444444445e-06, | |
| "loss": 0.1207, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 4.48, | |
| "learning_rate": 4.94888888888889e-06, | |
| "loss": 0.1215, | |
| "step": 2775 | |
| }, | |
| { | |
| "epoch": 4.52, | |
| "learning_rate": 4.893333333333334e-06, | |
| "loss": 0.1307, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 4.56, | |
| "learning_rate": 4.837777777777778e-06, | |
| "loss": 0.1187, | |
| "step": 2825 | |
| }, | |
| { | |
| "epoch": 4.6, | |
| "learning_rate": 4.7822222222222226e-06, | |
| "loss": 0.1388, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 4.64, | |
| "learning_rate": 4.7266666666666674e-06, | |
| "loss": 0.1355, | |
| "step": 2875 | |
| }, | |
| { | |
| "epoch": 4.68, | |
| "learning_rate": 4.6711111111111115e-06, | |
| "loss": 0.1482, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 4.72, | |
| "learning_rate": 4.6155555555555555e-06, | |
| "loss": 0.1242, | |
| "step": 2925 | |
| }, | |
| { | |
| "epoch": 4.77, | |
| "learning_rate": 4.56e-06, | |
| "loss": 0.1383, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 4.81, | |
| "learning_rate": 4.504444444444444e-06, | |
| "loss": 0.1301, | |
| "step": 2975 | |
| }, | |
| { | |
| "epoch": 4.85, | |
| "learning_rate": 4.448888888888889e-06, | |
| "loss": 0.1285, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 4.85, | |
| "eval_loss": 0.33927443623542786, | |
| "eval_runtime": 2631.0613, | |
| "eval_samples_per_second": 4.022, | |
| "eval_steps_per_second": 0.252, | |
| "eval_wer": 70.0375841272616, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 4.89, | |
| "learning_rate": 4.393333333333334e-06, | |
| "loss": 0.1417, | |
| "step": 3025 | |
| }, | |
| { | |
| "epoch": 4.93, | |
| "learning_rate": 4.337777777777778e-06, | |
| "loss": 0.1321, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 4.97, | |
| "learning_rate": 4.282222222222222e-06, | |
| "loss": 0.1149, | |
| "step": 3075 | |
| }, | |
| { | |
| "epoch": 5.01, | |
| "learning_rate": 4.226666666666667e-06, | |
| "loss": 0.1093, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 5.05, | |
| "learning_rate": 4.171111111111111e-06, | |
| "loss": 0.0887, | |
| "step": 3125 | |
| }, | |
| { | |
| "epoch": 5.09, | |
| "learning_rate": 4.115555555555556e-06, | |
| "loss": 0.0858, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 5.13, | |
| "learning_rate": 4.060000000000001e-06, | |
| "loss": 0.1015, | |
| "step": 3175 | |
| }, | |
| { | |
| "epoch": 5.17, | |
| "learning_rate": 4.004444444444445e-06, | |
| "loss": 0.0922, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 5.21, | |
| "learning_rate": 3.948888888888889e-06, | |
| "loss": 0.0821, | |
| "step": 3225 | |
| }, | |
| { | |
| "epoch": 5.25, | |
| "learning_rate": 3.893333333333333e-06, | |
| "loss": 0.0833, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 5.29, | |
| "learning_rate": 3.837777777777778e-06, | |
| "loss": 0.0897, | |
| "step": 3275 | |
| }, | |
| { | |
| "epoch": 5.33, | |
| "learning_rate": 3.782222222222223e-06, | |
| "loss": 0.1073, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 5.37, | |
| "learning_rate": 3.726666666666667e-06, | |
| "loss": 0.0869, | |
| "step": 3325 | |
| }, | |
| { | |
| "epoch": 5.41, | |
| "learning_rate": 3.6711111111111113e-06, | |
| "loss": 0.0903, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 5.45, | |
| "learning_rate": 3.615555555555556e-06, | |
| "loss": 0.101, | |
| "step": 3375 | |
| }, | |
| { | |
| "epoch": 5.49, | |
| "learning_rate": 3.5600000000000002e-06, | |
| "loss": 0.0845, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 5.53, | |
| "learning_rate": 3.5044444444444447e-06, | |
| "loss": 0.113, | |
| "step": 3425 | |
| }, | |
| { | |
| "epoch": 5.57, | |
| "learning_rate": 3.4488888888888896e-06, | |
| "loss": 0.0922, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 5.61, | |
| "learning_rate": 3.3933333333333336e-06, | |
| "loss": 0.0844, | |
| "step": 3475 | |
| }, | |
| { | |
| "epoch": 5.65, | |
| "learning_rate": 3.337777777777778e-06, | |
| "loss": 0.097, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 5.69, | |
| "learning_rate": 3.282222222222223e-06, | |
| "loss": 0.0836, | |
| "step": 3525 | |
| }, | |
| { | |
| "epoch": 5.73, | |
| "learning_rate": 3.226666666666667e-06, | |
| "loss": 0.1022, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 5.77, | |
| "learning_rate": 3.1711111111111114e-06, | |
| "loss": 0.0982, | |
| "step": 3575 | |
| }, | |
| { | |
| "epoch": 5.82, | |
| "learning_rate": 3.1155555555555555e-06, | |
| "loss": 0.0913, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 5.86, | |
| "learning_rate": 3.0600000000000003e-06, | |
| "loss": 0.1038, | |
| "step": 3625 | |
| }, | |
| { | |
| "epoch": 5.9, | |
| "learning_rate": 3.004444444444445e-06, | |
| "loss": 0.0898, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 5.94, | |
| "learning_rate": 2.948888888888889e-06, | |
| "loss": 0.1029, | |
| "step": 3675 | |
| }, | |
| { | |
| "epoch": 5.98, | |
| "learning_rate": 2.8933333333333337e-06, | |
| "loss": 0.101, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 6.02, | |
| "learning_rate": 2.837777777777778e-06, | |
| "loss": 0.0936, | |
| "step": 3725 | |
| }, | |
| { | |
| "epoch": 6.06, | |
| "learning_rate": 2.7822222222222222e-06, | |
| "loss": 0.0682, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 6.1, | |
| "learning_rate": 2.726666666666667e-06, | |
| "loss": 0.0752, | |
| "step": 3775 | |
| }, | |
| { | |
| "epoch": 6.14, | |
| "learning_rate": 2.6711111111111116e-06, | |
| "loss": 0.0693, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 6.18, | |
| "learning_rate": 2.6155555555555556e-06, | |
| "loss": 0.0662, | |
| "step": 3825 | |
| }, | |
| { | |
| "epoch": 6.22, | |
| "learning_rate": 2.56e-06, | |
| "loss": 0.0656, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 6.26, | |
| "learning_rate": 2.504444444444445e-06, | |
| "loss": 0.0684, | |
| "step": 3875 | |
| }, | |
| { | |
| "epoch": 6.3, | |
| "learning_rate": 2.448888888888889e-06, | |
| "loss": 0.06, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 6.34, | |
| "learning_rate": 2.3933333333333334e-06, | |
| "loss": 0.0604, | |
| "step": 3925 | |
| }, | |
| { | |
| "epoch": 6.38, | |
| "learning_rate": 2.337777777777778e-06, | |
| "loss": 0.0726, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 6.42, | |
| "learning_rate": 2.2822222222222223e-06, | |
| "loss": 0.0632, | |
| "step": 3975 | |
| }, | |
| { | |
| "epoch": 6.46, | |
| "learning_rate": 2.226666666666667e-06, | |
| "loss": 0.0706, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 6.46, | |
| "eval_loss": 0.3588755428791046, | |
| "eval_runtime": 2664.8672, | |
| "eval_samples_per_second": 3.971, | |
| "eval_steps_per_second": 0.248, | |
| "eval_wer": 70.41342539987764, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 6.5, | |
| "learning_rate": 2.1711111111111113e-06, | |
| "loss": 0.0721, | |
| "step": 4025 | |
| }, | |
| { | |
| "epoch": 6.54, | |
| "learning_rate": 2.1155555555555557e-06, | |
| "loss": 0.0693, | |
| "step": 4050 | |
| }, | |
| { | |
| "epoch": 6.58, | |
| "learning_rate": 2.06e-06, | |
| "loss": 0.0643, | |
| "step": 4075 | |
| }, | |
| { | |
| "epoch": 6.62, | |
| "learning_rate": 2.0044444444444446e-06, | |
| "loss": 0.0728, | |
| "step": 4100 | |
| }, | |
| { | |
| "epoch": 6.66, | |
| "learning_rate": 1.948888888888889e-06, | |
| "loss": 0.0704, | |
| "step": 4125 | |
| }, | |
| { | |
| "epoch": 6.7, | |
| "learning_rate": 1.8933333333333333e-06, | |
| "loss": 0.0691, | |
| "step": 4150 | |
| }, | |
| { | |
| "epoch": 6.74, | |
| "learning_rate": 1.837777777777778e-06, | |
| "loss": 0.0651, | |
| "step": 4175 | |
| }, | |
| { | |
| "epoch": 6.78, | |
| "learning_rate": 1.7822222222222225e-06, | |
| "loss": 0.0777, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 6.82, | |
| "learning_rate": 1.7266666666666667e-06, | |
| "loss": 0.0663, | |
| "step": 4225 | |
| }, | |
| { | |
| "epoch": 6.87, | |
| "learning_rate": 1.6711111111111112e-06, | |
| "loss": 0.0691, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 6.91, | |
| "learning_rate": 1.6155555555555559e-06, | |
| "loss": 0.0741, | |
| "step": 4275 | |
| }, | |
| { | |
| "epoch": 6.95, | |
| "learning_rate": 1.56e-06, | |
| "loss": 0.0744, | |
| "step": 4300 | |
| }, | |
| { | |
| "epoch": 6.99, | |
| "learning_rate": 1.5044444444444446e-06, | |
| "loss": 0.0722, | |
| "step": 4325 | |
| }, | |
| { | |
| "epoch": 7.03, | |
| "learning_rate": 1.4488888888888892e-06, | |
| "loss": 0.0567, | |
| "step": 4350 | |
| }, | |
| { | |
| "epoch": 7.07, | |
| "learning_rate": 1.3933333333333335e-06, | |
| "loss": 0.0558, | |
| "step": 4375 | |
| }, | |
| { | |
| "epoch": 7.11, | |
| "learning_rate": 1.337777777777778e-06, | |
| "loss": 0.0547, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 7.15, | |
| "learning_rate": 1.2822222222222222e-06, | |
| "loss": 0.0532, | |
| "step": 4425 | |
| }, | |
| { | |
| "epoch": 7.19, | |
| "learning_rate": 1.2266666666666666e-06, | |
| "loss": 0.0645, | |
| "step": 4450 | |
| }, | |
| { | |
| "epoch": 7.23, | |
| "learning_rate": 1.171111111111111e-06, | |
| "loss": 0.0413, | |
| "step": 4475 | |
| }, | |
| { | |
| "epoch": 7.27, | |
| "learning_rate": 1.1155555555555558e-06, | |
| "loss": 0.0468, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 7.31, | |
| "learning_rate": 1.06e-06, | |
| "loss": 0.0565, | |
| "step": 4525 | |
| }, | |
| { | |
| "epoch": 7.35, | |
| "learning_rate": 1.0044444444444445e-06, | |
| "loss": 0.0517, | |
| "step": 4550 | |
| }, | |
| { | |
| "epoch": 7.39, | |
| "learning_rate": 9.488888888888889e-07, | |
| "loss": 0.0535, | |
| "step": 4575 | |
| }, | |
| { | |
| "epoch": 7.43, | |
| "learning_rate": 8.933333333333334e-07, | |
| "loss": 0.0543, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 7.47, | |
| "learning_rate": 8.37777777777778e-07, | |
| "loss": 0.0474, | |
| "step": 4625 | |
| }, | |
| { | |
| "epoch": 7.51, | |
| "learning_rate": 7.822222222222223e-07, | |
| "loss": 0.0511, | |
| "step": 4650 | |
| }, | |
| { | |
| "epoch": 7.55, | |
| "learning_rate": 7.266666666666668e-07, | |
| "loss": 0.0569, | |
| "step": 4675 | |
| }, | |
| { | |
| "epoch": 7.59, | |
| "learning_rate": 6.711111111111111e-07, | |
| "loss": 0.0531, | |
| "step": 4700 | |
| }, | |
| { | |
| "epoch": 7.63, | |
| "learning_rate": 6.155555555555556e-07, | |
| "loss": 0.0464, | |
| "step": 4725 | |
| }, | |
| { | |
| "epoch": 7.67, | |
| "learning_rate": 5.6e-07, | |
| "loss": 0.0563, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 7.71, | |
| "learning_rate": 5.044444444444445e-07, | |
| "loss": 0.0527, | |
| "step": 4775 | |
| }, | |
| { | |
| "epoch": 7.75, | |
| "learning_rate": 4.488888888888889e-07, | |
| "loss": 0.0559, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 7.79, | |
| "learning_rate": 3.9333333333333336e-07, | |
| "loss": 0.0545, | |
| "step": 4825 | |
| }, | |
| { | |
| "epoch": 7.83, | |
| "learning_rate": 3.3777777777777777e-07, | |
| "loss": 0.0483, | |
| "step": 4850 | |
| }, | |
| { | |
| "epoch": 7.87, | |
| "learning_rate": 2.822222222222222e-07, | |
| "loss": 0.0549, | |
| "step": 4875 | |
| }, | |
| { | |
| "epoch": 7.92, | |
| "learning_rate": 2.266666666666667e-07, | |
| "loss": 0.059, | |
| "step": 4900 | |
| }, | |
| { | |
| "epoch": 7.96, | |
| "learning_rate": 1.7111111111111114e-07, | |
| "loss": 0.0562, | |
| "step": 4925 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "learning_rate": 1.1555555555555556e-07, | |
| "loss": 0.0571, | |
| "step": 4950 | |
| }, | |
| { | |
| "epoch": 8.04, | |
| "learning_rate": 6.000000000000001e-08, | |
| "loss": 0.0443, | |
| "step": 4975 | |
| }, | |
| { | |
| "epoch": 8.08, | |
| "learning_rate": 4.444444444444445e-09, | |
| "loss": 0.0484, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 8.08, | |
| "eval_loss": 0.36695703864097595, | |
| "eval_runtime": 2633.1605, | |
| "eval_samples_per_second": 4.018, | |
| "eval_steps_per_second": 0.251, | |
| "eval_wer": 70.65815925181366, | |
| "step": 5000 | |
| } | |
| ], | |
| "max_steps": 5000, | |
| "num_train_epochs": 9, | |
| "total_flos": 9.239581044965376e+19, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |