| { | |
| "best_metric": 64.95398773006134, | |
| "best_model_checkpoint": "./CN-whisper-th-medium_TH/checkpoint-4000", | |
| "epoch": 8.0, | |
| "global_step": 4000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 3.8e-07, | |
| "loss": 1.8059, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 8.8e-07, | |
| "loss": 1.2344, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.3800000000000001e-06, | |
| "loss": 0.9642, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8800000000000002e-06, | |
| "loss": 0.8623, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 2.38e-06, | |
| "loss": 0.65, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 2.88e-06, | |
| "loss": 0.6951, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 3.3800000000000007e-06, | |
| "loss": 0.4472, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 3.88e-06, | |
| "loss": 0.2373, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 4.38e-06, | |
| "loss": 0.0803, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 4.880000000000001e-06, | |
| "loss": 0.0745, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 5.380000000000001e-06, | |
| "loss": 0.0693, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 5.8800000000000005e-06, | |
| "loss": 0.0791, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 6.380000000000001e-06, | |
| "loss": 0.0815, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 6.88e-06, | |
| "loss": 0.1285, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 7.3800000000000005e-06, | |
| "loss": 0.1003, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 7.88e-06, | |
| "loss": 0.0998, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 8.380000000000001e-06, | |
| "loss": 0.1451, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 8.880000000000001e-06, | |
| "loss": 0.095, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 9.38e-06, | |
| "loss": 0.1281, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 9.88e-06, | |
| "loss": 0.1036, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 9.945714285714286e-06, | |
| "loss": 0.0978, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 9.874285714285715e-06, | |
| "loss": 0.1034, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 9.802857142857144e-06, | |
| "loss": 0.0949, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 9.731428571428573e-06, | |
| "loss": 0.1137, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 9.66e-06, | |
| "loss": 0.1113, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 9.58857142857143e-06, | |
| "loss": 0.1019, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 9.517142857142859e-06, | |
| "loss": 0.1204, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 9.445714285714288e-06, | |
| "loss": 0.091, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 9.374285714285715e-06, | |
| "loss": 0.136, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 9.302857142857144e-06, | |
| "loss": 0.143, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 9.231428571428573e-06, | |
| "loss": 0.1156, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 9.16e-06, | |
| "loss": 0.1047, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "learning_rate": 9.08857142857143e-06, | |
| "loss": 0.0884, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 9.017142857142858e-06, | |
| "loss": 0.085, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 8.945714285714286e-06, | |
| "loss": 0.1143, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 8.874285714285715e-06, | |
| "loss": 0.0963, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 8.802857142857144e-06, | |
| "loss": 0.1302, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 8.731428571428571e-06, | |
| "loss": 0.1024, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 8.66e-06, | |
| "loss": 0.0988, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 8.588571428571429e-06, | |
| "loss": 0.1031, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_loss": 0.2629605233669281, | |
| "eval_runtime": 223.3416, | |
| "eval_samples_per_second": 4.477, | |
| "eval_steps_per_second": 1.119, | |
| "eval_wer": 73.38957055214725, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 8.517142857142858e-06, | |
| "loss": 0.0608, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 8.445714285714285e-06, | |
| "loss": 0.0478, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 8.374285714285714e-06, | |
| "loss": 0.0552, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 8.302857142857143e-06, | |
| "loss": 0.0779, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 8.231428571428572e-06, | |
| "loss": 0.0661, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 8.16e-06, | |
| "loss": 0.0632, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 2.35, | |
| "learning_rate": 8.088571428571429e-06, | |
| "loss": 0.0641, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 8.017142857142858e-06, | |
| "loss": 0.0657, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 2.45, | |
| "learning_rate": 7.945714285714287e-06, | |
| "loss": 0.0744, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 7.874285714285716e-06, | |
| "loss": 0.0744, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 7.802857142857143e-06, | |
| "loss": 0.0621, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "learning_rate": 7.731428571428572e-06, | |
| "loss": 0.0556, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 2.65, | |
| "learning_rate": 7.660000000000001e-06, | |
| "loss": 0.0466, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 2.7, | |
| "learning_rate": 7.588571428571429e-06, | |
| "loss": 0.0608, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 7.5171428571428575e-06, | |
| "loss": 0.0449, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "learning_rate": 7.445714285714286e-06, | |
| "loss": 0.0663, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 2.85, | |
| "learning_rate": 7.374285714285715e-06, | |
| "loss": 0.0772, | |
| "step": 1425 | |
| }, | |
| { | |
| "epoch": 2.9, | |
| "learning_rate": 7.302857142857144e-06, | |
| "loss": 0.0629, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 2.95, | |
| "learning_rate": 7.231428571428573e-06, | |
| "loss": 0.0634, | |
| "step": 1475 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 7.16e-06, | |
| "loss": 0.0693, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 7.088571428571429e-06, | |
| "loss": 0.0181, | |
| "step": 1525 | |
| }, | |
| { | |
| "epoch": 3.1, | |
| "learning_rate": 7.017142857142858e-06, | |
| "loss": 0.0344, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 3.15, | |
| "learning_rate": 6.945714285714287e-06, | |
| "loss": 0.0333, | |
| "step": 1575 | |
| }, | |
| { | |
| "epoch": 3.2, | |
| "learning_rate": 6.874285714285714e-06, | |
| "loss": 0.0446, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 3.25, | |
| "learning_rate": 6.8028571428571434e-06, | |
| "loss": 0.0516, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 3.3, | |
| "learning_rate": 6.7314285714285724e-06, | |
| "loss": 0.0492, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 3.35, | |
| "learning_rate": 6.660000000000001e-06, | |
| "loss": 0.0245, | |
| "step": 1675 | |
| }, | |
| { | |
| "epoch": 3.4, | |
| "learning_rate": 6.588571428571429e-06, | |
| "loss": 0.0387, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 3.45, | |
| "learning_rate": 6.517142857142858e-06, | |
| "loss": 0.028, | |
| "step": 1725 | |
| }, | |
| { | |
| "epoch": 3.5, | |
| "learning_rate": 6.445714285714286e-06, | |
| "loss": 0.0221, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 3.55, | |
| "learning_rate": 6.374285714285715e-06, | |
| "loss": 0.0312, | |
| "step": 1775 | |
| }, | |
| { | |
| "epoch": 3.6, | |
| "learning_rate": 6.302857142857144e-06, | |
| "loss": 0.0353, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 3.65, | |
| "learning_rate": 6.231428571428571e-06, | |
| "loss": 0.0421, | |
| "step": 1825 | |
| }, | |
| { | |
| "epoch": 3.7, | |
| "learning_rate": 6.16e-06, | |
| "loss": 0.0189, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 3.75, | |
| "learning_rate": 6.088571428571429e-06, | |
| "loss": 0.0315, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 3.8, | |
| "learning_rate": 6.017142857142858e-06, | |
| "loss": 0.0205, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 3.85, | |
| "learning_rate": 5.945714285714286e-06, | |
| "loss": 0.0222, | |
| "step": 1925 | |
| }, | |
| { | |
| "epoch": 3.9, | |
| "learning_rate": 5.874285714285715e-06, | |
| "loss": 0.0376, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 3.95, | |
| "learning_rate": 5.802857142857144e-06, | |
| "loss": 0.0255, | |
| "step": 1975 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 5.731428571428572e-06, | |
| "loss": 0.03, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_loss": 0.2783181071281433, | |
| "eval_runtime": 224.4276, | |
| "eval_samples_per_second": 4.456, | |
| "eval_steps_per_second": 1.114, | |
| "eval_wer": 71.16564417177914, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 4.05, | |
| "learning_rate": 5.66e-06, | |
| "loss": 0.0224, | |
| "step": 2025 | |
| }, | |
| { | |
| "epoch": 4.1, | |
| "learning_rate": 5.588571428571429e-06, | |
| "loss": 0.0108, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 4.15, | |
| "learning_rate": 5.517142857142857e-06, | |
| "loss": 0.0132, | |
| "step": 2075 | |
| }, | |
| { | |
| "epoch": 4.2, | |
| "learning_rate": 5.445714285714286e-06, | |
| "loss": 0.0098, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 4.25, | |
| "learning_rate": 5.374285714285715e-06, | |
| "loss": 0.0128, | |
| "step": 2125 | |
| }, | |
| { | |
| "epoch": 4.3, | |
| "learning_rate": 5.3028571428571425e-06, | |
| "loss": 0.0055, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 4.35, | |
| "learning_rate": 5.2314285714285716e-06, | |
| "loss": 0.0105, | |
| "step": 2175 | |
| }, | |
| { | |
| "epoch": 4.4, | |
| "learning_rate": 5.1600000000000006e-06, | |
| "loss": 0.0116, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 4.45, | |
| "learning_rate": 5.08857142857143e-06, | |
| "loss": 0.016, | |
| "step": 2225 | |
| }, | |
| { | |
| "epoch": 4.5, | |
| "learning_rate": 5.017142857142857e-06, | |
| "loss": 0.0145, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 4.55, | |
| "learning_rate": 4.945714285714286e-06, | |
| "loss": 0.0132, | |
| "step": 2275 | |
| }, | |
| { | |
| "epoch": 4.6, | |
| "learning_rate": 4.874285714285715e-06, | |
| "loss": 0.0231, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 4.65, | |
| "learning_rate": 4.802857142857143e-06, | |
| "loss": 0.0157, | |
| "step": 2325 | |
| }, | |
| { | |
| "epoch": 4.7, | |
| "learning_rate": 4.731428571428572e-06, | |
| "loss": 0.0041, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 4.75, | |
| "learning_rate": 4.66e-06, | |
| "loss": 0.0073, | |
| "step": 2375 | |
| }, | |
| { | |
| "epoch": 4.8, | |
| "learning_rate": 4.588571428571429e-06, | |
| "loss": 0.0159, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 4.85, | |
| "learning_rate": 4.5171428571428575e-06, | |
| "loss": 0.0133, | |
| "step": 2425 | |
| }, | |
| { | |
| "epoch": 4.9, | |
| "learning_rate": 4.445714285714286e-06, | |
| "loss": 0.0124, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 4.95, | |
| "learning_rate": 4.374285714285715e-06, | |
| "loss": 0.0142, | |
| "step": 2475 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "learning_rate": 4.302857142857143e-06, | |
| "loss": 0.0142, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 5.05, | |
| "learning_rate": 4.231428571428572e-06, | |
| "loss": 0.0065, | |
| "step": 2525 | |
| }, | |
| { | |
| "epoch": 5.1, | |
| "learning_rate": 4.16e-06, | |
| "loss": 0.0026, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 5.15, | |
| "learning_rate": 4.088571428571429e-06, | |
| "loss": 0.0173, | |
| "step": 2575 | |
| }, | |
| { | |
| "epoch": 5.2, | |
| "learning_rate": 4.017142857142857e-06, | |
| "loss": 0.0092, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 5.25, | |
| "learning_rate": 3.945714285714286e-06, | |
| "loss": 0.0121, | |
| "step": 2625 | |
| }, | |
| { | |
| "epoch": 5.3, | |
| "learning_rate": 3.874285714285715e-06, | |
| "loss": 0.0005, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 5.35, | |
| "learning_rate": 3.802857142857143e-06, | |
| "loss": 0.0025, | |
| "step": 2675 | |
| }, | |
| { | |
| "epoch": 5.4, | |
| "learning_rate": 3.731428571428572e-06, | |
| "loss": 0.0069, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 5.45, | |
| "learning_rate": 3.66e-06, | |
| "loss": 0.0061, | |
| "step": 2725 | |
| }, | |
| { | |
| "epoch": 5.5, | |
| "learning_rate": 3.588571428571429e-06, | |
| "loss": 0.0031, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 5.55, | |
| "learning_rate": 3.5171428571428573e-06, | |
| "loss": 0.0091, | |
| "step": 2775 | |
| }, | |
| { | |
| "epoch": 5.6, | |
| "learning_rate": 3.4457142857142863e-06, | |
| "loss": 0.0099, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 5.65, | |
| "learning_rate": 3.3742857142857145e-06, | |
| "loss": 0.0063, | |
| "step": 2825 | |
| }, | |
| { | |
| "epoch": 5.7, | |
| "learning_rate": 3.302857142857143e-06, | |
| "loss": 0.0044, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 5.75, | |
| "learning_rate": 3.2314285714285716e-06, | |
| "loss": 0.006, | |
| "step": 2875 | |
| }, | |
| { | |
| "epoch": 5.8, | |
| "learning_rate": 3.1600000000000002e-06, | |
| "loss": 0.0061, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 5.85, | |
| "learning_rate": 3.0885714285714284e-06, | |
| "loss": 0.0034, | |
| "step": 2925 | |
| }, | |
| { | |
| "epoch": 5.9, | |
| "learning_rate": 3.0171428571428574e-06, | |
| "loss": 0.0068, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 5.95, | |
| "learning_rate": 2.945714285714286e-06, | |
| "loss": 0.003, | |
| "step": 2975 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "learning_rate": 2.8742857142857146e-06, | |
| "loss": 0.0042, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_loss": 0.2900189459323883, | |
| "eval_runtime": 225.1609, | |
| "eval_samples_per_second": 4.441, | |
| "eval_steps_per_second": 1.11, | |
| "eval_wer": 66.87116564417178, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 6.05, | |
| "learning_rate": 2.802857142857143e-06, | |
| "loss": 0.0048, | |
| "step": 3025 | |
| }, | |
| { | |
| "epoch": 6.1, | |
| "learning_rate": 2.7314285714285714e-06, | |
| "loss": 0.0002, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 6.15, | |
| "learning_rate": 2.6600000000000004e-06, | |
| "loss": 0.0003, | |
| "step": 3075 | |
| }, | |
| { | |
| "epoch": 6.2, | |
| "learning_rate": 2.5885714285714285e-06, | |
| "loss": 0.002, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 6.25, | |
| "learning_rate": 2.5171428571428575e-06, | |
| "loss": 0.0002, | |
| "step": 3125 | |
| }, | |
| { | |
| "epoch": 6.3, | |
| "learning_rate": 2.445714285714286e-06, | |
| "loss": 0.0012, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 6.35, | |
| "learning_rate": 2.3742857142857147e-06, | |
| "loss": 0.0007, | |
| "step": 3175 | |
| }, | |
| { | |
| "epoch": 6.4, | |
| "learning_rate": 2.302857142857143e-06, | |
| "loss": 0.0003, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 6.45, | |
| "learning_rate": 2.2314285714285715e-06, | |
| "loss": 0.0004, | |
| "step": 3225 | |
| }, | |
| { | |
| "epoch": 6.5, | |
| "learning_rate": 2.16e-06, | |
| "loss": 0.0002, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 6.55, | |
| "learning_rate": 2.0885714285714287e-06, | |
| "loss": 0.0015, | |
| "step": 3275 | |
| }, | |
| { | |
| "epoch": 6.6, | |
| "learning_rate": 2.0171428571428573e-06, | |
| "loss": 0.0006, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 6.65, | |
| "learning_rate": 1.945714285714286e-06, | |
| "loss": 0.0012, | |
| "step": 3325 | |
| }, | |
| { | |
| "epoch": 6.7, | |
| "learning_rate": 1.8742857142857142e-06, | |
| "loss": 0.0011, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 6.75, | |
| "learning_rate": 1.8028571428571432e-06, | |
| "loss": 0.0012, | |
| "step": 3375 | |
| }, | |
| { | |
| "epoch": 6.8, | |
| "learning_rate": 1.7314285714285716e-06, | |
| "loss": 0.0052, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 6.85, | |
| "learning_rate": 1.6600000000000002e-06, | |
| "loss": 0.0031, | |
| "step": 3425 | |
| }, | |
| { | |
| "epoch": 6.9, | |
| "learning_rate": 1.5885714285714288e-06, | |
| "loss": 0.0005, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 6.95, | |
| "learning_rate": 1.5171428571428574e-06, | |
| "loss": 0.0008, | |
| "step": 3475 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "learning_rate": 1.4457142857142858e-06, | |
| "loss": 0.0007, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 7.05, | |
| "learning_rate": 1.3742857142857143e-06, | |
| "loss": 0.0027, | |
| "step": 3525 | |
| }, | |
| { | |
| "epoch": 7.1, | |
| "learning_rate": 1.302857142857143e-06, | |
| "loss": 0.0002, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 7.15, | |
| "learning_rate": 1.2314285714285715e-06, | |
| "loss": 0.0001, | |
| "step": 3575 | |
| }, | |
| { | |
| "epoch": 7.2, | |
| "learning_rate": 1.1600000000000001e-06, | |
| "loss": 0.0003, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 7.25, | |
| "learning_rate": 1.0885714285714287e-06, | |
| "loss": 0.0001, | |
| "step": 3625 | |
| }, | |
| { | |
| "epoch": 7.3, | |
| "learning_rate": 1.0171428571428573e-06, | |
| "loss": 0.0001, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 7.35, | |
| "learning_rate": 9.457142857142858e-07, | |
| "loss": 0.0001, | |
| "step": 3675 | |
| }, | |
| { | |
| "epoch": 7.4, | |
| "learning_rate": 8.742857142857144e-07, | |
| "loss": 0.0037, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 7.45, | |
| "learning_rate": 8.028571428571429e-07, | |
| "loss": 0.0002, | |
| "step": 3725 | |
| }, | |
| { | |
| "epoch": 7.5, | |
| "learning_rate": 7.314285714285715e-07, | |
| "loss": 0.0001, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 7.55, | |
| "learning_rate": 6.6e-07, | |
| "loss": 0.0001, | |
| "step": 3775 | |
| }, | |
| { | |
| "epoch": 7.6, | |
| "learning_rate": 5.885714285714286e-07, | |
| "loss": 0.0001, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 7.65, | |
| "learning_rate": 5.171428571428572e-07, | |
| "loss": 0.0001, | |
| "step": 3825 | |
| }, | |
| { | |
| "epoch": 7.7, | |
| "learning_rate": 4.457142857142858e-07, | |
| "loss": 0.0001, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 7.75, | |
| "learning_rate": 3.7428571428571434e-07, | |
| "loss": 0.0001, | |
| "step": 3875 | |
| }, | |
| { | |
| "epoch": 7.8, | |
| "learning_rate": 3.028571428571429e-07, | |
| "loss": 0.0001, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 7.85, | |
| "learning_rate": 2.3142857142857144e-07, | |
| "loss": 0.0001, | |
| "step": 3925 | |
| }, | |
| { | |
| "epoch": 7.9, | |
| "learning_rate": 1.6e-07, | |
| "loss": 0.0001, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 7.95, | |
| "learning_rate": 8.857142857142858e-08, | |
| "loss": 0.0001, | |
| "step": 3975 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "learning_rate": 1.7142857142857143e-08, | |
| "loss": 0.0001, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_loss": 0.30384302139282227, | |
| "eval_runtime": 225.1306, | |
| "eval_samples_per_second": 4.442, | |
| "eval_steps_per_second": 1.11, | |
| "eval_wer": 64.95398773006134, | |
| "step": 4000 | |
| } | |
| ], | |
| "max_steps": 4000, | |
| "num_train_epochs": 8, | |
| "total_flos": 1.632967852032e+19, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |