| { | |
| "best_metric": 114.04808262084146, | |
| "best_model_checkpoint": "./whisper-small-hi-gpu_ft_latest/checkpoint-1000", | |
| "epoch": 19.51219512195122, | |
| "eval_steps": 1000, | |
| "global_step": 4000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 5.000000000000001e-07, | |
| "loss": 1.9792, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.0000000000000002e-06, | |
| "loss": 1.5051, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.5e-06, | |
| "loss": 1.0351, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 2.0000000000000003e-06, | |
| "loss": 0.7532, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 2.5e-06, | |
| "loss": 0.6512, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3e-06, | |
| "loss": 0.5881, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 3.5e-06, | |
| "loss": 0.5402, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 4.000000000000001e-06, | |
| "loss": 0.4919, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 4.5e-06, | |
| "loss": 0.4317, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 5e-06, | |
| "loss": 0.3807, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 5.500000000000001e-06, | |
| "loss": 0.3356, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 6e-06, | |
| "loss": 0.2905, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 6.5000000000000004e-06, | |
| "loss": 0.2465, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 7e-06, | |
| "loss": 0.221, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 7.500000000000001e-06, | |
| "loss": 0.2153, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 8.000000000000001e-06, | |
| "loss": 0.2057, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 8.5e-06, | |
| "loss": 0.1757, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 9e-06, | |
| "loss": 0.1332, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 2.32, | |
| "learning_rate": 9.5e-06, | |
| "loss": 0.1429, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 1e-05, | |
| "loss": 0.1396, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "learning_rate": 9.92857142857143e-06, | |
| "loss": 0.1367, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 2.68, | |
| "learning_rate": 9.857142857142859e-06, | |
| "loss": 0.1367, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "learning_rate": 9.785714285714286e-06, | |
| "loss": 0.1324, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 2.93, | |
| "learning_rate": 9.714285714285715e-06, | |
| "loss": 0.1249, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 9.642857142857144e-06, | |
| "loss": 0.1109, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 3.17, | |
| "learning_rate": 9.571428571428573e-06, | |
| "loss": 0.0717, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 3.29, | |
| "learning_rate": 9.5e-06, | |
| "loss": 0.0713, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 3.41, | |
| "learning_rate": 9.42857142857143e-06, | |
| "loss": 0.0684, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 3.54, | |
| "learning_rate": 9.357142857142859e-06, | |
| "loss": 0.0754, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 3.66, | |
| "learning_rate": 9.285714285714288e-06, | |
| "loss": 0.0723, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 3.78, | |
| "learning_rate": 9.214285714285715e-06, | |
| "loss": 0.0707, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 3.9, | |
| "learning_rate": 9.142857142857144e-06, | |
| "loss": 0.0725, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 4.02, | |
| "learning_rate": 9.071428571428573e-06, | |
| "loss": 0.0664, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 4.15, | |
| "learning_rate": 9e-06, | |
| "loss": 0.0389, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 4.27, | |
| "learning_rate": 8.92857142857143e-06, | |
| "loss": 0.0385, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 4.39, | |
| "learning_rate": 8.857142857142858e-06, | |
| "loss": 0.0367, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 4.51, | |
| "learning_rate": 8.785714285714286e-06, | |
| "loss": 0.0394, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 4.63, | |
| "learning_rate": 8.714285714285715e-06, | |
| "loss": 0.0382, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 4.76, | |
| "learning_rate": 8.642857142857144e-06, | |
| "loss": 0.0398, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 4.88, | |
| "learning_rate": 8.571428571428571e-06, | |
| "loss": 0.0365, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 4.88, | |
| "eval_loss": 0.3377941846847534, | |
| "eval_runtime": 8440.3955, | |
| "eval_samples_per_second": 0.343, | |
| "eval_steps_per_second": 0.021, | |
| "eval_wer": 114.04808262084146, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "learning_rate": 8.5e-06, | |
| "loss": 0.0365, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 5.12, | |
| "learning_rate": 8.428571428571429e-06, | |
| "loss": 0.0196, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 5.24, | |
| "learning_rate": 8.357142857142858e-06, | |
| "loss": 0.0223, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 5.37, | |
| "learning_rate": 8.285714285714287e-06, | |
| "loss": 0.0213, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 5.49, | |
| "learning_rate": 8.214285714285714e-06, | |
| "loss": 0.0205, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 5.61, | |
| "learning_rate": 8.142857142857143e-06, | |
| "loss": 0.0206, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 5.73, | |
| "learning_rate": 8.071428571428572e-06, | |
| "loss": 0.0193, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 5.85, | |
| "learning_rate": 8.000000000000001e-06, | |
| "loss": 0.0222, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 5.98, | |
| "learning_rate": 7.928571428571429e-06, | |
| "loss": 0.0209, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 6.1, | |
| "learning_rate": 7.857142857142858e-06, | |
| "loss": 0.0148, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 6.22, | |
| "learning_rate": 7.785714285714287e-06, | |
| "loss": 0.0118, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 6.34, | |
| "learning_rate": 7.714285714285716e-06, | |
| "loss": 0.0112, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 6.46, | |
| "learning_rate": 7.642857142857143e-06, | |
| "loss": 0.0112, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 6.59, | |
| "learning_rate": 7.571428571428572e-06, | |
| "loss": 0.0125, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 6.71, | |
| "learning_rate": 7.500000000000001e-06, | |
| "loss": 0.0124, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 6.83, | |
| "learning_rate": 7.428571428571429e-06, | |
| "loss": 0.0116, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 6.95, | |
| "learning_rate": 7.357142857142858e-06, | |
| "loss": 0.0113, | |
| "step": 1425 | |
| }, | |
| { | |
| "epoch": 7.07, | |
| "learning_rate": 7.285714285714286e-06, | |
| "loss": 0.0091, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 7.2, | |
| "learning_rate": 7.2142857142857145e-06, | |
| "loss": 0.0062, | |
| "step": 1475 | |
| }, | |
| { | |
| "epoch": 7.32, | |
| "learning_rate": 7.1428571428571436e-06, | |
| "loss": 0.0073, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 7.44, | |
| "learning_rate": 7.0714285714285726e-06, | |
| "loss": 0.0066, | |
| "step": 1525 | |
| }, | |
| { | |
| "epoch": 7.56, | |
| "learning_rate": 7e-06, | |
| "loss": 0.0059, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 7.68, | |
| "learning_rate": 6.928571428571429e-06, | |
| "loss": 0.0059, | |
| "step": 1575 | |
| }, | |
| { | |
| "epoch": 7.8, | |
| "learning_rate": 6.857142857142858e-06, | |
| "loss": 0.0065, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 7.93, | |
| "learning_rate": 6.785714285714287e-06, | |
| "loss": 0.0074, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 8.05, | |
| "learning_rate": 6.714285714285714e-06, | |
| "loss": 0.0051, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 8.17, | |
| "learning_rate": 6.642857142857143e-06, | |
| "loss": 0.004, | |
| "step": 1675 | |
| }, | |
| { | |
| "epoch": 8.29, | |
| "learning_rate": 6.571428571428572e-06, | |
| "loss": 0.004, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 8.41, | |
| "learning_rate": 6.5000000000000004e-06, | |
| "loss": 0.0032, | |
| "step": 1725 | |
| }, | |
| { | |
| "epoch": 8.54, | |
| "learning_rate": 6.4285714285714295e-06, | |
| "loss": 0.0032, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 8.66, | |
| "learning_rate": 6.357142857142858e-06, | |
| "loss": 0.0028, | |
| "step": 1775 | |
| }, | |
| { | |
| "epoch": 8.78, | |
| "learning_rate": 6.285714285714286e-06, | |
| "loss": 0.0038, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 8.9, | |
| "learning_rate": 6.214285714285715e-06, | |
| "loss": 0.0046, | |
| "step": 1825 | |
| }, | |
| { | |
| "epoch": 9.02, | |
| "learning_rate": 6.142857142857144e-06, | |
| "loss": 0.0036, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 9.15, | |
| "learning_rate": 6.071428571428571e-06, | |
| "loss": 0.0025, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 9.27, | |
| "learning_rate": 6e-06, | |
| "loss": 0.0027, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 9.39, | |
| "learning_rate": 5.928571428571429e-06, | |
| "loss": 0.002, | |
| "step": 1925 | |
| }, | |
| { | |
| "epoch": 9.51, | |
| "learning_rate": 5.857142857142858e-06, | |
| "loss": 0.0025, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 9.63, | |
| "learning_rate": 5.785714285714286e-06, | |
| "loss": 0.0018, | |
| "step": 1975 | |
| }, | |
| { | |
| "epoch": 9.76, | |
| "learning_rate": 5.7142857142857145e-06, | |
| "loss": 0.002, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 9.76, | |
| "eval_loss": 0.43459856510162354, | |
| "eval_runtime": 8851.1506, | |
| "eval_samples_per_second": 0.327, | |
| "eval_steps_per_second": 0.02, | |
| "eval_wer": 114.36976212647083, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 9.88, | |
| "learning_rate": 5.6428571428571435e-06, | |
| "loss": 0.0018, | |
| "step": 2025 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "learning_rate": 5.571428571428572e-06, | |
| "loss": 0.0022, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 10.12, | |
| "learning_rate": 5.500000000000001e-06, | |
| "loss": 0.0013, | |
| "step": 2075 | |
| }, | |
| { | |
| "epoch": 10.24, | |
| "learning_rate": 5.428571428571429e-06, | |
| "loss": 0.0014, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 10.37, | |
| "learning_rate": 5.357142857142857e-06, | |
| "loss": 0.001, | |
| "step": 2125 | |
| }, | |
| { | |
| "epoch": 10.49, | |
| "learning_rate": 5.285714285714286e-06, | |
| "loss": 0.0011, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 10.61, | |
| "learning_rate": 5.214285714285715e-06, | |
| "loss": 0.0009, | |
| "step": 2175 | |
| }, | |
| { | |
| "epoch": 10.73, | |
| "learning_rate": 5.142857142857142e-06, | |
| "loss": 0.001, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 10.85, | |
| "learning_rate": 5.071428571428571e-06, | |
| "loss": 0.0012, | |
| "step": 2225 | |
| }, | |
| { | |
| "epoch": 10.98, | |
| "learning_rate": 5e-06, | |
| "loss": 0.0011, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 11.1, | |
| "learning_rate": 4.928571428571429e-06, | |
| "loss": 0.0009, | |
| "step": 2275 | |
| }, | |
| { | |
| "epoch": 11.22, | |
| "learning_rate": 4.857142857142858e-06, | |
| "loss": 0.0008, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 11.34, | |
| "learning_rate": 4.785714285714287e-06, | |
| "loss": 0.0005, | |
| "step": 2325 | |
| }, | |
| { | |
| "epoch": 11.46, | |
| "learning_rate": 4.714285714285715e-06, | |
| "loss": 0.0005, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 11.59, | |
| "learning_rate": 4.642857142857144e-06, | |
| "loss": 0.0004, | |
| "step": 2375 | |
| }, | |
| { | |
| "epoch": 11.71, | |
| "learning_rate": 4.571428571428572e-06, | |
| "loss": 0.0005, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 11.83, | |
| "learning_rate": 4.5e-06, | |
| "loss": 0.0008, | |
| "step": 2425 | |
| }, | |
| { | |
| "epoch": 11.95, | |
| "learning_rate": 4.428571428571429e-06, | |
| "loss": 0.0006, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 12.07, | |
| "learning_rate": 4.357142857142857e-06, | |
| "loss": 0.0009, | |
| "step": 2475 | |
| }, | |
| { | |
| "epoch": 12.2, | |
| "learning_rate": 4.2857142857142855e-06, | |
| "loss": 0.0003, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 12.32, | |
| "learning_rate": 4.2142857142857145e-06, | |
| "loss": 0.0003, | |
| "step": 2525 | |
| }, | |
| { | |
| "epoch": 12.44, | |
| "learning_rate": 4.1428571428571435e-06, | |
| "loss": 0.0004, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 12.56, | |
| "learning_rate": 4.071428571428572e-06, | |
| "loss": 0.0003, | |
| "step": 2575 | |
| }, | |
| { | |
| "epoch": 12.68, | |
| "learning_rate": 4.000000000000001e-06, | |
| "loss": 0.0003, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 12.8, | |
| "learning_rate": 3.928571428571429e-06, | |
| "loss": 0.0003, | |
| "step": 2625 | |
| }, | |
| { | |
| "epoch": 12.93, | |
| "learning_rate": 3.857142857142858e-06, | |
| "loss": 0.0003, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 13.05, | |
| "learning_rate": 3.785714285714286e-06, | |
| "loss": 0.0003, | |
| "step": 2675 | |
| }, | |
| { | |
| "epoch": 13.17, | |
| "learning_rate": 3.7142857142857146e-06, | |
| "loss": 0.0002, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 13.29, | |
| "learning_rate": 3.642857142857143e-06, | |
| "loss": 0.0002, | |
| "step": 2725 | |
| }, | |
| { | |
| "epoch": 13.41, | |
| "learning_rate": 3.5714285714285718e-06, | |
| "loss": 0.0002, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 13.54, | |
| "learning_rate": 3.5e-06, | |
| "loss": 0.0002, | |
| "step": 2775 | |
| }, | |
| { | |
| "epoch": 13.66, | |
| "learning_rate": 3.428571428571429e-06, | |
| "loss": 0.0002, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 13.78, | |
| "learning_rate": 3.357142857142857e-06, | |
| "loss": 0.0003, | |
| "step": 2825 | |
| }, | |
| { | |
| "epoch": 13.9, | |
| "learning_rate": 3.285714285714286e-06, | |
| "loss": 0.0002, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 14.02, | |
| "learning_rate": 3.2142857142857147e-06, | |
| "loss": 0.0002, | |
| "step": 2875 | |
| }, | |
| { | |
| "epoch": 14.15, | |
| "learning_rate": 3.142857142857143e-06, | |
| "loss": 0.0002, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 14.27, | |
| "learning_rate": 3.071428571428572e-06, | |
| "loss": 0.0002, | |
| "step": 2925 | |
| }, | |
| { | |
| "epoch": 14.39, | |
| "learning_rate": 3e-06, | |
| "loss": 0.0002, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 14.51, | |
| "learning_rate": 2.928571428571429e-06, | |
| "loss": 0.0002, | |
| "step": 2975 | |
| }, | |
| { | |
| "epoch": 14.63, | |
| "learning_rate": 2.8571428571428573e-06, | |
| "loss": 0.0002, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 14.63, | |
| "eval_loss": 0.48673728108406067, | |
| "eval_runtime": 9299.1926, | |
| "eval_samples_per_second": 0.311, | |
| "eval_steps_per_second": 0.019, | |
| "eval_wer": 116.109371031914, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 14.76, | |
| "learning_rate": 2.785714285714286e-06, | |
| "loss": 0.0002, | |
| "step": 3025 | |
| }, | |
| { | |
| "epoch": 14.88, | |
| "learning_rate": 2.7142857142857144e-06, | |
| "loss": 0.0002, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 15.0, | |
| "learning_rate": 2.642857142857143e-06, | |
| "loss": 0.0002, | |
| "step": 3075 | |
| }, | |
| { | |
| "epoch": 15.12, | |
| "learning_rate": 2.571428571428571e-06, | |
| "loss": 0.0002, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 15.24, | |
| "learning_rate": 2.5e-06, | |
| "loss": 0.0002, | |
| "step": 3125 | |
| }, | |
| { | |
| "epoch": 15.37, | |
| "learning_rate": 2.428571428571429e-06, | |
| "loss": 0.0002, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 15.49, | |
| "learning_rate": 2.3571428571428574e-06, | |
| "loss": 0.0002, | |
| "step": 3175 | |
| }, | |
| { | |
| "epoch": 15.61, | |
| "learning_rate": 2.285714285714286e-06, | |
| "loss": 0.0002, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 15.73, | |
| "learning_rate": 2.2142857142857146e-06, | |
| "loss": 0.0002, | |
| "step": 3225 | |
| }, | |
| { | |
| "epoch": 15.85, | |
| "learning_rate": 2.1428571428571427e-06, | |
| "loss": 0.0002, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 15.98, | |
| "learning_rate": 2.0714285714285717e-06, | |
| "loss": 0.0002, | |
| "step": 3275 | |
| }, | |
| { | |
| "epoch": 16.1, | |
| "learning_rate": 2.0000000000000003e-06, | |
| "loss": 0.0002, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 16.22, | |
| "learning_rate": 1.928571428571429e-06, | |
| "loss": 0.0002, | |
| "step": 3325 | |
| }, | |
| { | |
| "epoch": 16.34, | |
| "learning_rate": 1.8571428571428573e-06, | |
| "loss": 0.0002, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 16.46, | |
| "learning_rate": 1.7857142857142859e-06, | |
| "loss": 0.0002, | |
| "step": 3375 | |
| }, | |
| { | |
| "epoch": 16.59, | |
| "learning_rate": 1.7142857142857145e-06, | |
| "loss": 0.0002, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 16.71, | |
| "learning_rate": 1.642857142857143e-06, | |
| "loss": 0.0002, | |
| "step": 3425 | |
| }, | |
| { | |
| "epoch": 16.83, | |
| "learning_rate": 1.5714285714285714e-06, | |
| "loss": 0.0001, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 16.95, | |
| "learning_rate": 1.5e-06, | |
| "loss": 0.0002, | |
| "step": 3475 | |
| }, | |
| { | |
| "epoch": 17.07, | |
| "learning_rate": 1.4285714285714286e-06, | |
| "loss": 0.0002, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 17.2, | |
| "learning_rate": 1.3571428571428572e-06, | |
| "loss": 0.0001, | |
| "step": 3525 | |
| }, | |
| { | |
| "epoch": 17.32, | |
| "learning_rate": 1.2857142857142856e-06, | |
| "loss": 0.0001, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 17.44, | |
| "learning_rate": 1.2142857142857144e-06, | |
| "loss": 0.0001, | |
| "step": 3575 | |
| }, | |
| { | |
| "epoch": 17.56, | |
| "learning_rate": 1.142857142857143e-06, | |
| "loss": 0.0001, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 17.68, | |
| "learning_rate": 1.0714285714285714e-06, | |
| "loss": 0.0001, | |
| "step": 3625 | |
| }, | |
| { | |
| "epoch": 17.8, | |
| "learning_rate": 1.0000000000000002e-06, | |
| "loss": 0.0001, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 17.93, | |
| "learning_rate": 9.285714285714287e-07, | |
| "loss": 0.0002, | |
| "step": 3675 | |
| }, | |
| { | |
| "epoch": 18.05, | |
| "learning_rate": 8.571428571428572e-07, | |
| "loss": 0.0002, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 18.17, | |
| "learning_rate": 7.857142857142857e-07, | |
| "loss": 0.0001, | |
| "step": 3725 | |
| }, | |
| { | |
| "epoch": 18.29, | |
| "learning_rate": 7.142857142857143e-07, | |
| "loss": 0.0001, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 18.41, | |
| "learning_rate": 6.428571428571428e-07, | |
| "loss": 0.0001, | |
| "step": 3775 | |
| }, | |
| { | |
| "epoch": 18.54, | |
| "learning_rate": 5.714285714285715e-07, | |
| "loss": 0.0001, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 18.66, | |
| "learning_rate": 5.000000000000001e-07, | |
| "loss": 0.0001, | |
| "step": 3825 | |
| }, | |
| { | |
| "epoch": 18.78, | |
| "learning_rate": 4.285714285714286e-07, | |
| "loss": 0.0001, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 18.9, | |
| "learning_rate": 3.5714285714285716e-07, | |
| "loss": 0.0001, | |
| "step": 3875 | |
| }, | |
| { | |
| "epoch": 19.02, | |
| "learning_rate": 2.8571428571428575e-07, | |
| "loss": 0.0001, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 19.15, | |
| "learning_rate": 2.142857142857143e-07, | |
| "loss": 0.0001, | |
| "step": 3925 | |
| }, | |
| { | |
| "epoch": 19.27, | |
| "learning_rate": 1.4285714285714287e-07, | |
| "loss": 0.0001, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 19.39, | |
| "learning_rate": 7.142857142857144e-08, | |
| "loss": 0.0001, | |
| "step": 3975 | |
| }, | |
| { | |
| "epoch": 19.51, | |
| "learning_rate": 0.0, | |
| "loss": 0.0001, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 19.51, | |
| "eval_loss": 0.5011982321739197, | |
| "eval_runtime": 9295.7529, | |
| "eval_samples_per_second": 0.311, | |
| "eval_steps_per_second": 0.019, | |
| "eval_wer": 117.1421315499873, | |
| "step": 4000 | |
| } | |
| ], | |
| "logging_steps": 25, | |
| "max_steps": 4000, | |
| "num_train_epochs": 20, | |
| "save_steps": 1000, | |
| "total_flos": 3.68292687888384e+19, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |