| { | |
| "best_metric": 109.13163064833007, | |
| "best_model_checkpoint": "./whisper-vietnamese-3/checkpoint-3000", | |
| "epoch": 21.58273381294964, | |
| "global_step": 3000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 4.6000000000000004e-07, | |
| "loss": 2.8534, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 9.600000000000001e-07, | |
| "loss": 2.0345, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.46e-06, | |
| "loss": 1.4058, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 1.9600000000000003e-06, | |
| "loss": 1.0909, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 2.46e-06, | |
| "loss": 0.889, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 2.96e-06, | |
| "loss": 0.7838, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 3.46e-06, | |
| "loss": 0.7152, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 3.96e-06, | |
| "loss": 0.6819, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 4.4600000000000005e-06, | |
| "loss": 0.6581, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 4.960000000000001e-06, | |
| "loss": 0.6276, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 5.460000000000001e-06, | |
| "loss": 0.6055, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 5.9600000000000005e-06, | |
| "loss": 0.5197, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 6.460000000000001e-06, | |
| "loss": 0.4698, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "learning_rate": 6.96e-06, | |
| "loss": 0.4506, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 2.7, | |
| "learning_rate": 7.4600000000000006e-06, | |
| "loss": 0.427, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 2.88, | |
| "learning_rate": 7.960000000000002e-06, | |
| "loss": 0.3566, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 3.06, | |
| "learning_rate": 8.46e-06, | |
| "loss": 0.3178, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 3.24, | |
| "learning_rate": 8.96e-06, | |
| "loss": 0.2256, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 3.42, | |
| "learning_rate": 9.460000000000001e-06, | |
| "loss": 0.2437, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 3.6, | |
| "learning_rate": 9.960000000000001e-06, | |
| "loss": 0.2447, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 3.78, | |
| "learning_rate": 9.934285714285715e-06, | |
| "loss": 0.2431, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 3.96, | |
| "learning_rate": 9.862857142857144e-06, | |
| "loss": 0.2328, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 4.14, | |
| "learning_rate": 9.791428571428571e-06, | |
| "loss": 0.1619, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 4.32, | |
| "learning_rate": 9.72e-06, | |
| "loss": 0.1336, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 4.5, | |
| "learning_rate": 9.648571428571429e-06, | |
| "loss": 0.1345, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 4.68, | |
| "learning_rate": 9.577142857142858e-06, | |
| "loss": 0.1369, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 4.86, | |
| "learning_rate": 9.505714285714287e-06, | |
| "loss": 0.1368, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 5.04, | |
| "learning_rate": 9.434285714285714e-06, | |
| "loss": 0.1359, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 5.22, | |
| "learning_rate": 9.362857142857143e-06, | |
| "loss": 0.0739, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 5.4, | |
| "learning_rate": 9.291428571428572e-06, | |
| "loss": 0.0754, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 5.58, | |
| "learning_rate": 9.220000000000002e-06, | |
| "loss": 0.0755, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 5.76, | |
| "learning_rate": 9.148571428571429e-06, | |
| "loss": 0.0732, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 5.94, | |
| "learning_rate": 9.077142857142858e-06, | |
| "loss": 0.07, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 6.12, | |
| "learning_rate": 9.005714285714287e-06, | |
| "loss": 0.0556, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 6.29, | |
| "learning_rate": 8.934285714285716e-06, | |
| "loss": 0.0374, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 6.47, | |
| "learning_rate": 8.862857142857143e-06, | |
| "loss": 0.0359, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 6.65, | |
| "learning_rate": 8.791428571428572e-06, | |
| "loss": 0.0392, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 6.83, | |
| "learning_rate": 8.720000000000001e-06, | |
| "loss": 0.0409, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 7.01, | |
| "learning_rate": 8.64857142857143e-06, | |
| "loss": 0.0402, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 7.19, | |
| "learning_rate": 8.577142857142858e-06, | |
| "loss": 0.0204, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 7.19, | |
| "eval_cer": 105.08606186100687, | |
| "eval_loss": 0.4899536669254303, | |
| "eval_runtime": 247.7316, | |
| "eval_samples_per_second": 0.997, | |
| "eval_steps_per_second": 0.125, | |
| "eval_wer": 123.61493123772102, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 7.37, | |
| "learning_rate": 8.505714285714287e-06, | |
| "loss": 0.0215, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 7.55, | |
| "learning_rate": 8.434285714285716e-06, | |
| "loss": 0.0214, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 7.73, | |
| "learning_rate": 8.362857142857143e-06, | |
| "loss": 0.0222, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 7.91, | |
| "learning_rate": 8.291428571428572e-06, | |
| "loss": 0.0234, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 8.09, | |
| "learning_rate": 8.220000000000001e-06, | |
| "loss": 0.016, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 8.27, | |
| "learning_rate": 8.148571428571428e-06, | |
| "loss": 0.0135, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 8.45, | |
| "learning_rate": 8.077142857142857e-06, | |
| "loss": 0.0144, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 8.63, | |
| "learning_rate": 8.005714285714286e-06, | |
| "loss": 0.0139, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 8.81, | |
| "learning_rate": 7.934285714285715e-06, | |
| "loss": 0.0126, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 8.99, | |
| "learning_rate": 7.862857142857143e-06, | |
| "loss": 0.0146, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 9.17, | |
| "learning_rate": 7.791428571428572e-06, | |
| "loss": 0.0089, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 9.35, | |
| "learning_rate": 7.72e-06, | |
| "loss": 0.0081, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 9.53, | |
| "learning_rate": 7.64857142857143e-06, | |
| "loss": 0.0094, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 9.71, | |
| "learning_rate": 7.577142857142857e-06, | |
| "loss": 0.0084, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 9.89, | |
| "learning_rate": 7.505714285714286e-06, | |
| "loss": 0.0075, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 10.07, | |
| "learning_rate": 7.434285714285715e-06, | |
| "loss": 0.0084, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 10.25, | |
| "learning_rate": 7.362857142857144e-06, | |
| "loss": 0.0063, | |
| "step": 1425 | |
| }, | |
| { | |
| "epoch": 10.43, | |
| "learning_rate": 7.291428571428571e-06, | |
| "loss": 0.0057, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 10.61, | |
| "learning_rate": 7.22e-06, | |
| "loss": 0.0055, | |
| "step": 1475 | |
| }, | |
| { | |
| "epoch": 10.79, | |
| "learning_rate": 7.148571428571429e-06, | |
| "loss": 0.0062, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 10.97, | |
| "learning_rate": 7.077142857142858e-06, | |
| "loss": 0.0068, | |
| "step": 1525 | |
| }, | |
| { | |
| "epoch": 11.15, | |
| "learning_rate": 7.0057142857142865e-06, | |
| "loss": 0.0041, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 11.33, | |
| "learning_rate": 6.934285714285715e-06, | |
| "loss": 0.0045, | |
| "step": 1575 | |
| }, | |
| { | |
| "epoch": 11.51, | |
| "learning_rate": 6.862857142857144e-06, | |
| "loss": 0.0047, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 11.69, | |
| "learning_rate": 6.791428571428572e-06, | |
| "loss": 0.0048, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 11.87, | |
| "learning_rate": 6.720000000000001e-06, | |
| "loss": 0.0037, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 12.05, | |
| "learning_rate": 6.648571428571429e-06, | |
| "loss": 0.0034, | |
| "step": 1675 | |
| }, | |
| { | |
| "epoch": 12.23, | |
| "learning_rate": 6.577142857142857e-06, | |
| "loss": 0.0031, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 12.41, | |
| "learning_rate": 6.505714285714286e-06, | |
| "loss": 0.003, | |
| "step": 1725 | |
| }, | |
| { | |
| "epoch": 12.59, | |
| "learning_rate": 6.434285714285715e-06, | |
| "loss": 0.0044, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 12.77, | |
| "learning_rate": 6.3628571428571426e-06, | |
| "loss": 0.0033, | |
| "step": 1775 | |
| }, | |
| { | |
| "epoch": 12.95, | |
| "learning_rate": 6.2914285714285716e-06, | |
| "loss": 0.0032, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 13.13, | |
| "learning_rate": 6.220000000000001e-06, | |
| "loss": 0.0032, | |
| "step": 1825 | |
| }, | |
| { | |
| "epoch": 13.31, | |
| "learning_rate": 6.14857142857143e-06, | |
| "loss": 0.0027, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 13.49, | |
| "learning_rate": 6.077142857142858e-06, | |
| "loss": 0.0029, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 13.67, | |
| "learning_rate": 6.005714285714286e-06, | |
| "loss": 0.0028, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 13.85, | |
| "learning_rate": 5.934285714285715e-06, | |
| "loss": 0.0023, | |
| "step": 1925 | |
| }, | |
| { | |
| "epoch": 14.03, | |
| "learning_rate": 5.862857142857143e-06, | |
| "loss": 0.0019, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 14.21, | |
| "learning_rate": 5.791428571428572e-06, | |
| "loss": 0.0029, | |
| "step": 1975 | |
| }, | |
| { | |
| "epoch": 14.39, | |
| "learning_rate": 5.72e-06, | |
| "loss": 0.0022, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 14.39, | |
| "eval_cer": 104.70151047347889, | |
| "eval_loss": 0.5671390295028687, | |
| "eval_runtime": 247.6486, | |
| "eval_samples_per_second": 0.997, | |
| "eval_steps_per_second": 0.125, | |
| "eval_wer": 111.96070726915521, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 14.57, | |
| "learning_rate": 5.6485714285714285e-06, | |
| "loss": 0.0027, | |
| "step": 2025 | |
| }, | |
| { | |
| "epoch": 14.75, | |
| "learning_rate": 5.5771428571428575e-06, | |
| "loss": 0.0024, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 14.93, | |
| "learning_rate": 5.5057142857142865e-06, | |
| "loss": 0.0029, | |
| "step": 2075 | |
| }, | |
| { | |
| "epoch": 15.11, | |
| "learning_rate": 5.4342857142857155e-06, | |
| "loss": 0.0016, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 15.29, | |
| "learning_rate": 5.362857142857143e-06, | |
| "loss": 0.0018, | |
| "step": 2125 | |
| }, | |
| { | |
| "epoch": 15.47, | |
| "learning_rate": 5.291428571428572e-06, | |
| "loss": 0.0018, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 15.65, | |
| "learning_rate": 5.220000000000001e-06, | |
| "loss": 0.0027, | |
| "step": 2175 | |
| }, | |
| { | |
| "epoch": 15.83, | |
| "learning_rate": 5.14857142857143e-06, | |
| "loss": 0.0015, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 16.01, | |
| "learning_rate": 5.077142857142857e-06, | |
| "loss": 0.0016, | |
| "step": 2225 | |
| }, | |
| { | |
| "epoch": 16.19, | |
| "learning_rate": 5.005714285714286e-06, | |
| "loss": 0.0013, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 16.37, | |
| "learning_rate": 4.934285714285715e-06, | |
| "loss": 0.002, | |
| "step": 2275 | |
| }, | |
| { | |
| "epoch": 16.55, | |
| "learning_rate": 4.862857142857143e-06, | |
| "loss": 0.0019, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 16.73, | |
| "learning_rate": 4.7914285714285715e-06, | |
| "loss": 0.0013, | |
| "step": 2325 | |
| }, | |
| { | |
| "epoch": 16.91, | |
| "learning_rate": 4.7200000000000005e-06, | |
| "loss": 0.0017, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 17.09, | |
| "learning_rate": 4.648571428571429e-06, | |
| "loss": 0.0015, | |
| "step": 2375 | |
| }, | |
| { | |
| "epoch": 17.27, | |
| "learning_rate": 4.577142857142858e-06, | |
| "loss": 0.0013, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 17.45, | |
| "learning_rate": 4.505714285714286e-06, | |
| "loss": 0.0017, | |
| "step": 2425 | |
| }, | |
| { | |
| "epoch": 17.63, | |
| "learning_rate": 4.434285714285715e-06, | |
| "loss": 0.0019, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 17.81, | |
| "learning_rate": 4.362857142857143e-06, | |
| "loss": 0.002, | |
| "step": 2475 | |
| }, | |
| { | |
| "epoch": 17.99, | |
| "learning_rate": 4.291428571428572e-06, | |
| "loss": 0.0017, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 18.17, | |
| "learning_rate": 4.22e-06, | |
| "loss": 0.001, | |
| "step": 2525 | |
| }, | |
| { | |
| "epoch": 18.35, | |
| "learning_rate": 4.148571428571429e-06, | |
| "loss": 0.0009, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 18.53, | |
| "learning_rate": 4.0771428571428574e-06, | |
| "loss": 0.0012, | |
| "step": 2575 | |
| }, | |
| { | |
| "epoch": 18.71, | |
| "learning_rate": 4.0057142857142864e-06, | |
| "loss": 0.0016, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 18.88, | |
| "learning_rate": 3.934285714285715e-06, | |
| "loss": 0.0018, | |
| "step": 2625 | |
| }, | |
| { | |
| "epoch": 19.06, | |
| "learning_rate": 3.862857142857143e-06, | |
| "loss": 0.0014, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 19.24, | |
| "learning_rate": 3.7914285714285722e-06, | |
| "loss": 0.0012, | |
| "step": 2675 | |
| }, | |
| { | |
| "epoch": 19.42, | |
| "learning_rate": 3.7200000000000004e-06, | |
| "loss": 0.001, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 19.6, | |
| "learning_rate": 3.648571428571429e-06, | |
| "loss": 0.0014, | |
| "step": 2725 | |
| }, | |
| { | |
| "epoch": 19.78, | |
| "learning_rate": 3.5771428571428576e-06, | |
| "loss": 0.0013, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 19.96, | |
| "learning_rate": 3.505714285714286e-06, | |
| "loss": 0.0009, | |
| "step": 2775 | |
| }, | |
| { | |
| "epoch": 20.14, | |
| "learning_rate": 3.4342857142857143e-06, | |
| "loss": 0.0008, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 20.32, | |
| "learning_rate": 3.3628571428571433e-06, | |
| "loss": 0.0016, | |
| "step": 2825 | |
| }, | |
| { | |
| "epoch": 20.5, | |
| "learning_rate": 3.2914285714285715e-06, | |
| "loss": 0.0008, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 20.68, | |
| "learning_rate": 3.2200000000000005e-06, | |
| "loss": 0.0014, | |
| "step": 2875 | |
| }, | |
| { | |
| "epoch": 20.86, | |
| "learning_rate": 3.1485714285714287e-06, | |
| "loss": 0.001, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 21.04, | |
| "learning_rate": 3.0771428571428573e-06, | |
| "loss": 0.0009, | |
| "step": 2925 | |
| }, | |
| { | |
| "epoch": 21.22, | |
| "learning_rate": 3.005714285714286e-06, | |
| "loss": 0.0007, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 21.4, | |
| "learning_rate": 2.9342857142857144e-06, | |
| "loss": 0.0011, | |
| "step": 2975 | |
| }, | |
| { | |
| "epoch": 21.58, | |
| "learning_rate": 2.8628571428571435e-06, | |
| "loss": 0.0009, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 21.58, | |
| "eval_cer": 101.53081033112092, | |
| "eval_loss": 0.6016765236854553, | |
| "eval_runtime": 253.4179, | |
| "eval_samples_per_second": 0.975, | |
| "eval_steps_per_second": 0.122, | |
| "eval_wer": 109.13163064833007, | |
| "step": 3000 | |
| } | |
| ], | |
| "max_steps": 4000, | |
| "num_train_epochs": 29, | |
| "total_flos": 1.380967716151296e+19, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |