| { | |
| "best_metric": 43.130896226415096, | |
| "best_model_checkpoint": "./whisper-small-bem\\checkpoint-4000", | |
| "epoch": 0.6440186765416197, | |
| "global_step": 4000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 5.000000000000001e-07, | |
| "loss": 5.0182, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.0000000000000002e-06, | |
| "loss": 4.199, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.5e-06, | |
| "loss": 3.3383, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 2.0000000000000003e-06, | |
| "loss": 2.6709, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 2.5e-06, | |
| "loss": 2.2656, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 3e-06, | |
| "loss": 2.1601, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 3.5e-06, | |
| "loss": 1.8449, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 4.000000000000001e-06, | |
| "loss": 1.7861, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 4.5e-06, | |
| "loss": 1.8245, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 5e-06, | |
| "loss": 1.4688, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 5.500000000000001e-06, | |
| "loss": 1.2695, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 6e-06, | |
| "loss": 1.4908, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 6.5000000000000004e-06, | |
| "loss": 1.1877, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 7e-06, | |
| "loss": 1.2925, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 7.500000000000001e-06, | |
| "loss": 1.1981, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 8.000000000000001e-06, | |
| "loss": 1.0733, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 8.5e-06, | |
| "loss": 1.0012, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9e-06, | |
| "loss": 0.9732, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.5e-06, | |
| "loss": 0.9498, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1e-05, | |
| "loss": 0.9393, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.92857142857143e-06, | |
| "loss": 0.905, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.857142857142859e-06, | |
| "loss": 0.9295, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.785714285714286e-06, | |
| "loss": 0.91, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.714285714285715e-06, | |
| "loss": 0.9162, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.642857142857144e-06, | |
| "loss": 0.8577, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.571428571428573e-06, | |
| "loss": 0.7675, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.5e-06, | |
| "loss": 0.8088, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.42857142857143e-06, | |
| "loss": 0.8569, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.357142857142859e-06, | |
| "loss": 0.7501, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.285714285714288e-06, | |
| "loss": 0.7226, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.214285714285715e-06, | |
| "loss": 0.8245, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 9.142857142857144e-06, | |
| "loss": 0.8411, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 9.071428571428573e-06, | |
| "loss": 0.8415, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9e-06, | |
| "loss": 0.7309, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 8.92857142857143e-06, | |
| "loss": 0.7156, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 8.857142857142858e-06, | |
| "loss": 0.8099, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 8.785714285714286e-06, | |
| "loss": 0.725, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 8.714285714285715e-06, | |
| "loss": 0.7525, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 8.642857142857144e-06, | |
| "loss": 0.6894, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 8.571428571428571e-06, | |
| "loss": 0.6962, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "eval_loss": 0.6828086972236633, | |
| "eval_runtime": 15569.5401, | |
| "eval_samples_per_second": 0.088, | |
| "eval_steps_per_second": 0.011, | |
| "eval_wer": 58.475825471698116, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 8.5e-06, | |
| "loss": 0.7957, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 8.428571428571429e-06, | |
| "loss": 0.668, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 8.357142857142858e-06, | |
| "loss": 0.6491, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 8.285714285714287e-06, | |
| "loss": 0.7405, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 8.214285714285714e-06, | |
| "loss": 0.7403, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 8.142857142857143e-06, | |
| "loss": 0.6591, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 8.071428571428572e-06, | |
| "loss": 0.7558, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 8.000000000000001e-06, | |
| "loss": 0.6271, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 7.928571428571429e-06, | |
| "loss": 0.6028, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 7.857142857142858e-06, | |
| "loss": 0.6532, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 7.785714285714287e-06, | |
| "loss": 0.7154, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 7.714285714285716e-06, | |
| "loss": 0.7344, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 7.642857142857143e-06, | |
| "loss": 0.6992, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 7.571428571428572e-06, | |
| "loss": 0.6866, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 7.500000000000001e-06, | |
| "loss": 0.6101, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 7.428571428571429e-06, | |
| "loss": 0.6792, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 7.357142857142858e-06, | |
| "loss": 0.647, | |
| "step": 1425 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 7.285714285714286e-06, | |
| "loss": 0.6041, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 7.2142857142857145e-06, | |
| "loss": 0.6981, | |
| "step": 1475 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 7.1428571428571436e-06, | |
| "loss": 0.6133, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 7.0714285714285726e-06, | |
| "loss": 0.613, | |
| "step": 1525 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 7e-06, | |
| "loss": 0.6198, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 6.928571428571429e-06, | |
| "loss": 0.6177, | |
| "step": 1575 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 6.857142857142858e-06, | |
| "loss": 0.5481, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 6.785714285714287e-06, | |
| "loss": 0.7077, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 6.714285714285714e-06, | |
| "loss": 0.657, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 6.642857142857143e-06, | |
| "loss": 0.7048, | |
| "step": 1675 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 6.571428571428572e-06, | |
| "loss": 0.7918, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 6.5000000000000004e-06, | |
| "loss": 0.6718, | |
| "step": 1725 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 6.4285714285714295e-06, | |
| "loss": 0.7668, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 6.357142857142858e-06, | |
| "loss": 0.6188, | |
| "step": 1775 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 6.285714285714286e-06, | |
| "loss": 0.6733, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 6.214285714285715e-06, | |
| "loss": 0.6764, | |
| "step": 1825 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 6.142857142857144e-06, | |
| "loss": 0.6698, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 6.071428571428571e-06, | |
| "loss": 0.5567, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 6e-06, | |
| "loss": 0.5529, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 5.928571428571429e-06, | |
| "loss": 0.5934, | |
| "step": 1925 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 5.857142857142858e-06, | |
| "loss": 0.546, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 5.785714285714286e-06, | |
| "loss": 0.5833, | |
| "step": 1975 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 5.7142857142857145e-06, | |
| "loss": 0.6228, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "eval_loss": 0.5770470499992371, | |
| "eval_runtime": 14396.5231, | |
| "eval_samples_per_second": 0.095, | |
| "eval_steps_per_second": 0.012, | |
| "eval_wer": 50.884433962264154, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 5.6428571428571435e-06, | |
| "loss": 0.5725, | |
| "step": 2025 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 5.571428571428572e-06, | |
| "loss": 0.5806, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 5.500000000000001e-06, | |
| "loss": 0.5848, | |
| "step": 2075 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 5.428571428571429e-06, | |
| "loss": 0.62, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 5.357142857142857e-06, | |
| "loss": 0.5509, | |
| "step": 2125 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 5.285714285714286e-06, | |
| "loss": 0.6506, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 5.214285714285715e-06, | |
| "loss": 0.5729, | |
| "step": 2175 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 5.142857142857142e-06, | |
| "loss": 0.5747, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 5.071428571428571e-06, | |
| "loss": 0.609, | |
| "step": 2225 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 5e-06, | |
| "loss": 0.6445, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 4.928571428571429e-06, | |
| "loss": 0.6285, | |
| "step": 2275 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 4.857142857142858e-06, | |
| "loss": 0.6245, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 4.785714285714287e-06, | |
| "loss": 0.5406, | |
| "step": 2325 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 4.714285714285715e-06, | |
| "loss": 0.6286, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 4.642857142857144e-06, | |
| "loss": 0.5853, | |
| "step": 2375 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 4.571428571428572e-06, | |
| "loss": 0.4812, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 4.5e-06, | |
| "loss": 0.6213, | |
| "step": 2425 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 4.428571428571429e-06, | |
| "loss": 0.5938, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 4.357142857142857e-06, | |
| "loss": 0.4531, | |
| "step": 2475 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 4.2857142857142855e-06, | |
| "loss": 0.6101, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 4.2142857142857145e-06, | |
| "loss": 0.6078, | |
| "step": 2525 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 4.1428571428571435e-06, | |
| "loss": 0.5647, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 4.071428571428572e-06, | |
| "loss": 0.6066, | |
| "step": 2575 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 4.000000000000001e-06, | |
| "loss": 0.4813, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 3.928571428571429e-06, | |
| "loss": 0.5535, | |
| "step": 2625 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 3.857142857142858e-06, | |
| "loss": 0.6569, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 3.785714285714286e-06, | |
| "loss": 0.4314, | |
| "step": 2675 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 3.7142857142857146e-06, | |
| "loss": 0.5806, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 3.642857142857143e-06, | |
| "loss": 0.6602, | |
| "step": 2725 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 3.5714285714285718e-06, | |
| "loss": 0.6429, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 3.5e-06, | |
| "loss": 0.4359, | |
| "step": 2775 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 3.428571428571429e-06, | |
| "loss": 0.5956, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 3.357142857142857e-06, | |
| "loss": 0.5895, | |
| "step": 2825 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 3.285714285714286e-06, | |
| "loss": 0.4916, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 3.2142857142857147e-06, | |
| "loss": 0.5987, | |
| "step": 2875 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 3.142857142857143e-06, | |
| "loss": 0.4935, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 3.071428571428572e-06, | |
| "loss": 0.5933, | |
| "step": 2925 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 3e-06, | |
| "loss": 0.4789, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 2.928571428571429e-06, | |
| "loss": 0.5486, | |
| "step": 2975 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 2.8571428571428573e-06, | |
| "loss": 0.5414, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "eval_loss": 0.5073037147521973, | |
| "eval_runtime": 15231.0297, | |
| "eval_samples_per_second": 0.09, | |
| "eval_steps_per_second": 0.011, | |
| "eval_wer": 45.16509433962264, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 2.785714285714286e-06, | |
| "loss": 0.5552, | |
| "step": 3025 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 2.7142857142857144e-06, | |
| "loss": 0.5026, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 2.642857142857143e-06, | |
| "loss": 0.4991, | |
| "step": 3075 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 2.571428571428571e-06, | |
| "loss": 0.5121, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 2.5e-06, | |
| "loss": 0.5131, | |
| "step": 3125 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 2.428571428571429e-06, | |
| "loss": 0.6146, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 2.3571428571428574e-06, | |
| "loss": 0.4602, | |
| "step": 3175 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 2.285714285714286e-06, | |
| "loss": 0.5307, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 2.2142857142857146e-06, | |
| "loss": 0.5, | |
| "step": 3225 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 2.1428571428571427e-06, | |
| "loss": 0.4883, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 2.0714285714285717e-06, | |
| "loss": 0.4782, | |
| "step": 3275 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 2.0000000000000003e-06, | |
| "loss": 0.4218, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.928571428571429e-06, | |
| "loss": 0.5049, | |
| "step": 3325 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.8571428571428573e-06, | |
| "loss": 0.505, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.7857142857142859e-06, | |
| "loss": 0.407, | |
| "step": 3375 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 1.7142857142857145e-06, | |
| "loss": 0.4835, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 1.642857142857143e-06, | |
| "loss": 0.4616, | |
| "step": 3425 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.5714285714285714e-06, | |
| "loss": 0.6107, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.5e-06, | |
| "loss": 0.5041, | |
| "step": 3475 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.4285714285714286e-06, | |
| "loss": 0.4478, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.3571428571428572e-06, | |
| "loss": 0.4552, | |
| "step": 3525 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.2857142857142856e-06, | |
| "loss": 0.4953, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.2142857142857144e-06, | |
| "loss": 0.5477, | |
| "step": 3575 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.142857142857143e-06, | |
| "loss": 0.5025, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.0714285714285714e-06, | |
| "loss": 0.4985, | |
| "step": 3625 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 1.0000000000000002e-06, | |
| "loss": 0.4588, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 9.285714285714287e-07, | |
| "loss": 0.5554, | |
| "step": 3675 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 8.571428571428572e-07, | |
| "loss": 0.5152, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.857142857142857e-07, | |
| "loss": 0.5861, | |
| "step": 3725 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.142857142857143e-07, | |
| "loss": 0.5224, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 6.428571428571428e-07, | |
| "loss": 0.4235, | |
| "step": 3775 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 5.714285714285715e-07, | |
| "loss": 0.5631, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 5.000000000000001e-07, | |
| "loss": 0.5361, | |
| "step": 3825 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 4.285714285714286e-07, | |
| "loss": 0.5528, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.5714285714285716e-07, | |
| "loss": 0.4646, | |
| "step": 3875 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 2.8571428571428575e-07, | |
| "loss": 0.468, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 2.142857142857143e-07, | |
| "loss": 0.4007, | |
| "step": 3925 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.4285714285714287e-07, | |
| "loss": 0.4645, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 7.142857142857144e-08, | |
| "loss": 0.5833, | |
| "step": 3975 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.0, | |
| "loss": 0.4313, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "eval_loss": 0.4780045449733734, | |
| "eval_runtime": 15398.4722, | |
| "eval_samples_per_second": 0.089, | |
| "eval_steps_per_second": 0.011, | |
| "eval_wer": 43.130896226415096, | |
| "step": 4000 | |
| } | |
| ], | |
| "max_steps": 4000, | |
| "num_train_epochs": 1, | |
| "total_flos": 2.30868320256e+18, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |