| { | |
| "best_metric": 12.981869792143577, | |
| "best_model_checkpoint": "/data_drive_500/whisper_finetune/checkpoint-5000", | |
| "epoch": 2.1284, | |
| "global_step": 5000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 4.4e-07, | |
| "loss": 0.7346, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.400000000000001e-07, | |
| "loss": 0.6306, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.44e-06, | |
| "loss": 0.5211, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.94e-06, | |
| "loss": 0.392, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.4400000000000004e-06, | |
| "loss": 0.3631, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.9400000000000002e-06, | |
| "loss": 0.3985, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 3.44e-06, | |
| "loss": 0.3412, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 3.94e-06, | |
| "loss": 0.2767, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 4.440000000000001e-06, | |
| "loss": 0.3361, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 4.94e-06, | |
| "loss": 0.314, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 5.4400000000000004e-06, | |
| "loss": 0.2937, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 5.94e-06, | |
| "loss": 0.2849, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 6.440000000000001e-06, | |
| "loss": 0.3048, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 6.9400000000000005e-06, | |
| "loss": 0.2954, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 7.440000000000001e-06, | |
| "loss": 0.2716, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 7.94e-06, | |
| "loss": 0.2426, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 8.44e-06, | |
| "loss": 0.3007, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 8.94e-06, | |
| "loss": 0.221, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.440000000000001e-06, | |
| "loss": 0.2947, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.940000000000001e-06, | |
| "loss": 0.2706, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.951111111111111e-06, | |
| "loss": 0.2865, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.895555555555557e-06, | |
| "loss": 0.2953, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.84e-06, | |
| "loss": 0.2566, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.784444444444445e-06, | |
| "loss": 0.305, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.72888888888889e-06, | |
| "loss": 0.2996, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 9.673333333333334e-06, | |
| "loss": 0.2655, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.617777777777778e-06, | |
| "loss": 0.2858, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.562222222222223e-06, | |
| "loss": 0.2232, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.506666666666667e-06, | |
| "loss": 0.2653, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 9.451111111111112e-06, | |
| "loss": 0.2982, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 9.395555555555556e-06, | |
| "loss": 0.2542, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 9.340000000000002e-06, | |
| "loss": 0.295, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.284444444444444e-06, | |
| "loss": 0.3007, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.22888888888889e-06, | |
| "loss": 0.2876, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.173333333333334e-06, | |
| "loss": 0.2682, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 9.117777777777778e-06, | |
| "loss": 0.2863, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 9.062222222222224e-06, | |
| "loss": 0.2676, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 9.006666666666666e-06, | |
| "loss": 0.24, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 8.951111111111112e-06, | |
| "loss": 0.2535, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 8.895555555555556e-06, | |
| "loss": 0.218, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "eval_loss": 0.29696622490882874, | |
| "eval_runtime": 4337.2036, | |
| "eval_samples_per_second": 0.667, | |
| "eval_steps_per_second": 0.667, | |
| "eval_wer": 20.153832066660563, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 8.84e-06, | |
| "loss": 0.3025, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 8.784444444444446e-06, | |
| "loss": 0.2344, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 8.72888888888889e-06, | |
| "loss": 0.2608, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 8.673333333333334e-06, | |
| "loss": 0.2689, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 8.617777777777778e-06, | |
| "loss": 0.1799, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 8.562222222222224e-06, | |
| "loss": 0.1894, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 8.506666666666668e-06, | |
| "loss": 0.2221, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 8.451111111111112e-06, | |
| "loss": 0.2069, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 8.395555555555557e-06, | |
| "loss": 0.2259, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 8.34e-06, | |
| "loss": 0.2224, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 8.284444444444446e-06, | |
| "loss": 0.1824, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 8.22888888888889e-06, | |
| "loss": 0.2063, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 8.173333333333334e-06, | |
| "loss": 0.1975, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 8.11777777777778e-06, | |
| "loss": 0.1889, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 8.062222222222222e-06, | |
| "loss": 0.2216, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 8.006666666666667e-06, | |
| "loss": 0.1761, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 7.951111111111111e-06, | |
| "loss": 0.2305, | |
| "step": 1425 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 7.895555555555557e-06, | |
| "loss": 0.2258, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 7.840000000000001e-06, | |
| "loss": 0.1817, | |
| "step": 1475 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 7.784444444444445e-06, | |
| "loss": 0.1718, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 7.72888888888889e-06, | |
| "loss": 0.1859, | |
| "step": 1525 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 7.673333333333333e-06, | |
| "loss": 0.179, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 7.617777777777778e-06, | |
| "loss": 0.2325, | |
| "step": 1575 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 7.562222222222223e-06, | |
| "loss": 0.2054, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 7.506666666666668e-06, | |
| "loss": 0.1893, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 7.451111111111111e-06, | |
| "loss": 0.1958, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 7.395555555555556e-06, | |
| "loss": 0.1918, | |
| "step": 1675 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 7.340000000000001e-06, | |
| "loss": 0.1956, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 7.284444444444445e-06, | |
| "loss": 0.2003, | |
| "step": 1725 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 7.22888888888889e-06, | |
| "loss": 0.1553, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 7.173333333333335e-06, | |
| "loss": 0.1897, | |
| "step": 1775 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 7.117777777777778e-06, | |
| "loss": 0.1898, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 7.062222222222223e-06, | |
| "loss": 0.1709, | |
| "step": 1825 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 7.006666666666667e-06, | |
| "loss": 0.1809, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 6.951111111111112e-06, | |
| "loss": 0.1779, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 6.8955555555555565e-06, | |
| "loss": 0.15, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 6.8400000000000014e-06, | |
| "loss": 0.1691, | |
| "step": 1925 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 6.784444444444445e-06, | |
| "loss": 0.1436, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 6.7288888888888895e-06, | |
| "loss": 0.1482, | |
| "step": 1975 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 6.6733333333333335e-06, | |
| "loss": 0.1537, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "eval_loss": 0.2573223114013672, | |
| "eval_runtime": 4323.9005, | |
| "eval_samples_per_second": 0.669, | |
| "eval_steps_per_second": 0.669, | |
| "eval_wer": 17.253456643164544, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 6.617777777777778e-06, | |
| "loss": 0.171, | |
| "step": 2025 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 6.562222222222223e-06, | |
| "loss": 0.1811, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 6.5066666666666665e-06, | |
| "loss": 0.1903, | |
| "step": 2075 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 6.451111111111111e-06, | |
| "loss": 0.1751, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 6.395555555555556e-06, | |
| "loss": 0.1623, | |
| "step": 2125 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 6.34e-06, | |
| "loss": 0.1393, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 6.284444444444445e-06, | |
| "loss": 0.1286, | |
| "step": 2175 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 6.22888888888889e-06, | |
| "loss": 0.1521, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 6.173333333333333e-06, | |
| "loss": 0.1159, | |
| "step": 2225 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 6.117777777777778e-06, | |
| "loss": 0.1349, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 6.062222222222223e-06, | |
| "loss": 0.1348, | |
| "step": 2275 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 6.006666666666667e-06, | |
| "loss": 0.1513, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 5.951111111111112e-06, | |
| "loss": 0.1554, | |
| "step": 2325 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 5.895555555555557e-06, | |
| "loss": 0.1253, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 5.84e-06, | |
| "loss": 0.1368, | |
| "step": 2375 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 5.784444444444445e-06, | |
| "loss": 0.1391, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 5.72888888888889e-06, | |
| "loss": 0.1229, | |
| "step": 2425 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 5.673333333333334e-06, | |
| "loss": 0.1006, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 5.617777777777779e-06, | |
| "loss": 0.1084, | |
| "step": 2475 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 5.562222222222222e-06, | |
| "loss": 0.1321, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 5.506666666666667e-06, | |
| "loss": 0.091, | |
| "step": 2525 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 5.451111111111112e-06, | |
| "loss": 0.1158, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 5.3955555555555565e-06, | |
| "loss": 0.099, | |
| "step": 2575 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 5.3400000000000005e-06, | |
| "loss": 0.1051, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 5.2844444444444454e-06, | |
| "loss": 0.1005, | |
| "step": 2625 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 5.228888888888889e-06, | |
| "loss": 0.0895, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 5.1733333333333335e-06, | |
| "loss": 0.0762, | |
| "step": 2675 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 5.117777777777778e-06, | |
| "loss": 0.0978, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 5.062222222222222e-06, | |
| "loss": 0.0902, | |
| "step": 2725 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 5.006666666666667e-06, | |
| "loss": 0.0994, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 4.951111111111111e-06, | |
| "loss": 0.1231, | |
| "step": 2775 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 4.895555555555556e-06, | |
| "loss": 0.0944, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 4.84e-06, | |
| "loss": 0.064, | |
| "step": 2825 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 4.784444444444445e-06, | |
| "loss": 0.0752, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 4.728888888888889e-06, | |
| "loss": 0.0918, | |
| "step": 2875 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 4.673333333333333e-06, | |
| "loss": 0.1143, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 4.617777777777778e-06, | |
| "loss": 0.0834, | |
| "step": 2925 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 4.562222222222222e-06, | |
| "loss": 0.1155, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 4.506666666666667e-06, | |
| "loss": 0.0932, | |
| "step": 2975 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 4.451111111111112e-06, | |
| "loss": 0.0802, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "eval_loss": 0.23921515047550201, | |
| "eval_runtime": 4291.5122, | |
| "eval_samples_per_second": 0.674, | |
| "eval_steps_per_second": 0.674, | |
| "eval_wer": 14.27982785459207, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 4.395555555555556e-06, | |
| "loss": 0.1002, | |
| "step": 3025 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 4.34e-06, | |
| "loss": 0.0561, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 4.284444444444445e-06, | |
| "loss": 0.0922, | |
| "step": 3075 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 4.228888888888889e-06, | |
| "loss": 0.0978, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 4.173333333333334e-06, | |
| "loss": 0.0871, | |
| "step": 3125 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 4.117777777777779e-06, | |
| "loss": 0.076, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 4.062222222222223e-06, | |
| "loss": 0.0849, | |
| "step": 3175 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 4.006666666666667e-06, | |
| "loss": 0.0725, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 3.951111111111112e-06, | |
| "loss": 0.0768, | |
| "step": 3225 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 3.895555555555556e-06, | |
| "loss": 0.0619, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 3.8400000000000005e-06, | |
| "loss": 0.0948, | |
| "step": 3275 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 3.784444444444445e-06, | |
| "loss": 0.0918, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 3.728888888888889e-06, | |
| "loss": 0.0797, | |
| "step": 3325 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 3.673333333333334e-06, | |
| "loss": 0.0561, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 3.617777777777778e-06, | |
| "loss": 0.0483, | |
| "step": 3375 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 3.5622222222222224e-06, | |
| "loss": 0.0787, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 3.5066666666666673e-06, | |
| "loss": 0.0795, | |
| "step": 3425 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 3.4511111111111113e-06, | |
| "loss": 0.077, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 3.3955555555555558e-06, | |
| "loss": 0.0584, | |
| "step": 3475 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 3.3400000000000006e-06, | |
| "loss": 0.055, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 3.2844444444444447e-06, | |
| "loss": 0.0503, | |
| "step": 3525 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 3.228888888888889e-06, | |
| "loss": 0.0722, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 3.173333333333334e-06, | |
| "loss": 0.0493, | |
| "step": 3575 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 3.117777777777778e-06, | |
| "loss": 0.0771, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 3.0622222222222225e-06, | |
| "loss": 0.0684, | |
| "step": 3625 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 3.0066666666666674e-06, | |
| "loss": 0.0443, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 2.9511111111111114e-06, | |
| "loss": 0.0703, | |
| "step": 3675 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 2.895555555555556e-06, | |
| "loss": 0.0739, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 2.84e-06, | |
| "loss": 0.0605, | |
| "step": 3725 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 2.784444444444445e-06, | |
| "loss": 0.0706, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 2.7288888888888893e-06, | |
| "loss": 0.0642, | |
| "step": 3775 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 2.6733333333333333e-06, | |
| "loss": 0.0555, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 2.617777777777778e-06, | |
| "loss": 0.0576, | |
| "step": 3825 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 2.5622222222222226e-06, | |
| "loss": 0.0674, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 2.5066666666666667e-06, | |
| "loss": 0.0607, | |
| "step": 3875 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 2.451111111111111e-06, | |
| "loss": 0.0554, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 2.3955555555555556e-06, | |
| "loss": 0.0345, | |
| "step": 3925 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 2.3400000000000005e-06, | |
| "loss": 0.0561, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 2.2844444444444445e-06, | |
| "loss": 0.0628, | |
| "step": 3975 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 2.228888888888889e-06, | |
| "loss": 0.0521, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "eval_loss": 0.2262759655714035, | |
| "eval_runtime": 4281.7956, | |
| "eval_samples_per_second": 0.676, | |
| "eval_steps_per_second": 0.676, | |
| "eval_wer": 13.714403442908157, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 2.1733333333333334e-06, | |
| "loss": 0.0543, | |
| "step": 4025 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 2.117777777777778e-06, | |
| "loss": 0.0435, | |
| "step": 4050 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 2.0622222222222223e-06, | |
| "loss": 0.0636, | |
| "step": 4075 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 2.006666666666667e-06, | |
| "loss": 0.0616, | |
| "step": 4100 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 1.9511111111111113e-06, | |
| "loss": 0.0453, | |
| "step": 4125 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 1.8955555555555557e-06, | |
| "loss": 0.0443, | |
| "step": 4150 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 1.8422222222222225e-06, | |
| "loss": 0.038, | |
| "step": 4175 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 1.7866666666666668e-06, | |
| "loss": 0.0509, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 1.7311111111111112e-06, | |
| "loss": 0.0314, | |
| "step": 4225 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 1.675555555555556e-06, | |
| "loss": 0.0424, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 1.6200000000000002e-06, | |
| "loss": 0.0464, | |
| "step": 4275 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 1.5644444444444446e-06, | |
| "loss": 0.0566, | |
| "step": 4300 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 1.5088888888888889e-06, | |
| "loss": 0.0551, | |
| "step": 4325 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 1.4533333333333335e-06, | |
| "loss": 0.0523, | |
| "step": 4350 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 1.397777777777778e-06, | |
| "loss": 0.0478, | |
| "step": 4375 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 1.3422222222222222e-06, | |
| "loss": 0.0398, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 1.286666666666667e-06, | |
| "loss": 0.0316, | |
| "step": 4425 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 1.2311111111111112e-06, | |
| "loss": 0.0257, | |
| "step": 4450 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 1.1755555555555556e-06, | |
| "loss": 0.0367, | |
| "step": 4475 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 1.12e-06, | |
| "loss": 0.0485, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 1.0644444444444445e-06, | |
| "loss": 0.0506, | |
| "step": 4525 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 1.008888888888889e-06, | |
| "loss": 0.052, | |
| "step": 4550 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 9.533333333333335e-07, | |
| "loss": 0.0302, | |
| "step": 4575 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 8.977777777777778e-07, | |
| "loss": 0.0255, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 8.422222222222224e-07, | |
| "loss": 0.027, | |
| "step": 4625 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 7.866666666666667e-07, | |
| "loss": 0.0448, | |
| "step": 4650 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 7.311111111111112e-07, | |
| "loss": 0.0171, | |
| "step": 4675 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 6.755555555555555e-07, | |
| "loss": 0.0358, | |
| "step": 4700 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 6.200000000000001e-07, | |
| "loss": 0.0167, | |
| "step": 4725 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 5.644444444444445e-07, | |
| "loss": 0.0304, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 5.088888888888889e-07, | |
| "loss": 0.0218, | |
| "step": 4775 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 4.533333333333334e-07, | |
| "loss": 0.0266, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 3.9777777777777783e-07, | |
| "loss": 0.0267, | |
| "step": 4825 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 3.422222222222223e-07, | |
| "loss": 0.0309, | |
| "step": 4850 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 2.866666666666667e-07, | |
| "loss": 0.0281, | |
| "step": 4875 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 2.3111111111111112e-07, | |
| "loss": 0.0246, | |
| "step": 4900 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 1.7555555555555558e-07, | |
| "loss": 0.0323, | |
| "step": 4925 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 1.2000000000000002e-07, | |
| "loss": 0.0317, | |
| "step": 4950 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 6.444444444444445e-08, | |
| "loss": 0.0311, | |
| "step": 4975 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 8.88888888888889e-09, | |
| "loss": 0.0135, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "eval_loss": 0.23063671588897705, | |
| "eval_runtime": 4288.3264, | |
| "eval_samples_per_second": 0.675, | |
| "eval_steps_per_second": 0.675, | |
| "eval_wer": 12.981869792143577, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "step": 5000, | |
| "total_flos": 1.02060490752e+19, | |
| "train_loss": 0.1471910211980343, | |
| "train_runtime": 31362.9902, | |
| "train_samples_per_second": 0.319, | |
| "train_steps_per_second": 0.159 | |
| } | |
| ], | |
| "max_steps": 5000, | |
| "num_train_epochs": 9223372036854775807, | |
| "total_flos": 1.02060490752e+19, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |