|
{ |
|
"best_metric": 43.64254021075985, |
|
"best_model_checkpoint": "./whisper-tiny-it-5/checkpoint-3000", |
|
"epoch": 2.8625954198473282, |
|
"global_step": 3000, |
|
"is_hyper_param_search": false, |
|
"is_local_process_zero": true, |
|
"is_world_process_zero": true, |
|
"log_history": [ |
|
{ |
|
"epoch": 0.02, |
|
"learning_rate": 4.000000000000001e-06, |
|
"loss": 3.391, |
|
"step": 25 |
|
}, |
|
{ |
|
"epoch": 0.05, |
|
"learning_rate": 9e-06, |
|
"loss": 1.6624, |
|
"step": 50 |
|
}, |
|
{ |
|
"epoch": 0.07, |
|
"learning_rate": 1.4000000000000001e-05, |
|
"loss": 0.8609, |
|
"step": 75 |
|
}, |
|
{ |
|
"epoch": 0.1, |
|
"learning_rate": 1.9e-05, |
|
"loss": 0.8187, |
|
"step": 100 |
|
}, |
|
{ |
|
"epoch": 0.12, |
|
"learning_rate": 2.4e-05, |
|
"loss": 0.7349, |
|
"step": 125 |
|
}, |
|
{ |
|
"epoch": 0.14, |
|
"learning_rate": 2.9e-05, |
|
"loss": 0.7348, |
|
"step": 150 |
|
}, |
|
{ |
|
"epoch": 0.17, |
|
"learning_rate": 3.4000000000000007e-05, |
|
"loss": 0.7649, |
|
"step": 175 |
|
}, |
|
{ |
|
"epoch": 0.19, |
|
"learning_rate": 3.9000000000000006e-05, |
|
"loss": 0.7318, |
|
"step": 200 |
|
}, |
|
{ |
|
"epoch": 0.21, |
|
"learning_rate": 4.4000000000000006e-05, |
|
"loss": 0.6739, |
|
"step": 225 |
|
}, |
|
{ |
|
"epoch": 0.24, |
|
"learning_rate": 4.9e-05, |
|
"loss": 0.7161, |
|
"step": 250 |
|
}, |
|
{ |
|
"epoch": 0.26, |
|
"learning_rate": 5.4000000000000005e-05, |
|
"loss": 0.712, |
|
"step": 275 |
|
}, |
|
{ |
|
"epoch": 0.29, |
|
"learning_rate": 5.9e-05, |
|
"loss": 0.7173, |
|
"step": 300 |
|
}, |
|
{ |
|
"epoch": 0.31, |
|
"learning_rate": 6.400000000000001e-05, |
|
"loss": 0.7434, |
|
"step": 325 |
|
}, |
|
{ |
|
"epoch": 0.33, |
|
"learning_rate": 6.9e-05, |
|
"loss": 0.7158, |
|
"step": 350 |
|
}, |
|
{ |
|
"epoch": 0.36, |
|
"learning_rate": 7.4e-05, |
|
"loss": 0.7698, |
|
"step": 375 |
|
}, |
|
{ |
|
"epoch": 0.38, |
|
"learning_rate": 7.900000000000001e-05, |
|
"loss": 0.7483, |
|
"step": 400 |
|
}, |
|
{ |
|
"epoch": 0.41, |
|
"learning_rate": 8.4e-05, |
|
"loss": 0.7544, |
|
"step": 425 |
|
}, |
|
{ |
|
"epoch": 0.43, |
|
"learning_rate": 8.900000000000001e-05, |
|
"loss": 0.7746, |
|
"step": 450 |
|
}, |
|
{ |
|
"epoch": 0.45, |
|
"learning_rate": 9.4e-05, |
|
"loss": 0.7408, |
|
"step": 475 |
|
}, |
|
{ |
|
"epoch": 0.48, |
|
"learning_rate": 9.900000000000001e-05, |
|
"loss": 0.7807, |
|
"step": 500 |
|
}, |
|
{ |
|
"epoch": 0.5, |
|
"learning_rate": 9.942857142857144e-05, |
|
"loss": 0.8248, |
|
"step": 525 |
|
}, |
|
{ |
|
"epoch": 0.52, |
|
"learning_rate": 9.871428571428572e-05, |
|
"loss": 0.7842, |
|
"step": 550 |
|
}, |
|
{ |
|
"epoch": 0.55, |
|
"learning_rate": 9.8e-05, |
|
"loss": 0.8156, |
|
"step": 575 |
|
}, |
|
{ |
|
"epoch": 0.57, |
|
"learning_rate": 9.728571428571429e-05, |
|
"loss": 0.7807, |
|
"step": 600 |
|
}, |
|
{ |
|
"epoch": 0.6, |
|
"learning_rate": 9.657142857142858e-05, |
|
"loss": 0.8033, |
|
"step": 625 |
|
}, |
|
{ |
|
"epoch": 0.62, |
|
"learning_rate": 9.585714285714285e-05, |
|
"loss": 0.7965, |
|
"step": 650 |
|
}, |
|
{ |
|
"epoch": 0.64, |
|
"learning_rate": 9.514285714285714e-05, |
|
"loss": 0.7454, |
|
"step": 675 |
|
}, |
|
{ |
|
"epoch": 0.67, |
|
"learning_rate": 9.442857142857144e-05, |
|
"loss": 0.7695, |
|
"step": 700 |
|
}, |
|
{ |
|
"epoch": 0.69, |
|
"learning_rate": 9.371428571428572e-05, |
|
"loss": 0.7727, |
|
"step": 725 |
|
}, |
|
{ |
|
"epoch": 0.72, |
|
"learning_rate": 9.300000000000001e-05, |
|
"loss": 0.7356, |
|
"step": 750 |
|
}, |
|
{ |
|
"epoch": 0.74, |
|
"learning_rate": 9.228571428571429e-05, |
|
"loss": 0.7532, |
|
"step": 775 |
|
}, |
|
{ |
|
"epoch": 0.76, |
|
"learning_rate": 9.157142857142857e-05, |
|
"loss": 0.7034, |
|
"step": 800 |
|
}, |
|
{ |
|
"epoch": 0.79, |
|
"learning_rate": 9.085714285714286e-05, |
|
"loss": 0.7396, |
|
"step": 825 |
|
}, |
|
{ |
|
"epoch": 0.81, |
|
"learning_rate": 9.014285714285716e-05, |
|
"loss": 0.7645, |
|
"step": 850 |
|
}, |
|
{ |
|
"epoch": 0.83, |
|
"learning_rate": 8.942857142857142e-05, |
|
"loss": 0.7284, |
|
"step": 875 |
|
}, |
|
{ |
|
"epoch": 0.86, |
|
"learning_rate": 8.871428571428572e-05, |
|
"loss": 0.7457, |
|
"step": 900 |
|
}, |
|
{ |
|
"epoch": 0.88, |
|
"learning_rate": 8.800000000000001e-05, |
|
"loss": 0.7921, |
|
"step": 925 |
|
}, |
|
{ |
|
"epoch": 0.91, |
|
"learning_rate": 8.728571428571429e-05, |
|
"loss": 0.6815, |
|
"step": 950 |
|
}, |
|
{ |
|
"epoch": 0.93, |
|
"learning_rate": 8.657142857142858e-05, |
|
"loss": 0.7065, |
|
"step": 975 |
|
}, |
|
{ |
|
"epoch": 0.95, |
|
"learning_rate": 8.585714285714286e-05, |
|
"loss": 0.7015, |
|
"step": 1000 |
|
}, |
|
{ |
|
"epoch": 0.95, |
|
"eval_loss": 0.9463410377502441, |
|
"eval_runtime": 445.1552, |
|
"eval_samples_per_second": 3.37, |
|
"eval_steps_per_second": 0.422, |
|
"eval_wer": 64.46894065446477, |
|
"step": 1000 |
|
}, |
|
{ |
|
"epoch": 0.98, |
|
"learning_rate": 8.514285714285714e-05, |
|
"loss": 0.7318, |
|
"step": 1025 |
|
}, |
|
{ |
|
"epoch": 1.0, |
|
"learning_rate": 8.442857142857144e-05, |
|
"loss": 0.6619, |
|
"step": 1050 |
|
}, |
|
{ |
|
"epoch": 1.03, |
|
"learning_rate": 8.371428571428572e-05, |
|
"loss": 0.3724, |
|
"step": 1075 |
|
}, |
|
{ |
|
"epoch": 1.05, |
|
"learning_rate": 8.3e-05, |
|
"loss": 0.3719, |
|
"step": 1100 |
|
}, |
|
{ |
|
"epoch": 1.07, |
|
"learning_rate": 8.228571428571429e-05, |
|
"loss": 0.3858, |
|
"step": 1125 |
|
}, |
|
{ |
|
"epoch": 1.1, |
|
"learning_rate": 8.157142857142857e-05, |
|
"loss": 0.4176, |
|
"step": 1150 |
|
}, |
|
{ |
|
"epoch": 1.12, |
|
"learning_rate": 8.085714285714287e-05, |
|
"loss": 0.3893, |
|
"step": 1175 |
|
}, |
|
{ |
|
"epoch": 1.15, |
|
"learning_rate": 8.014285714285715e-05, |
|
"loss": 0.3824, |
|
"step": 1200 |
|
}, |
|
{ |
|
"epoch": 1.17, |
|
"learning_rate": 7.942857142857143e-05, |
|
"loss": 0.4098, |
|
"step": 1225 |
|
}, |
|
{ |
|
"epoch": 1.19, |
|
"learning_rate": 7.871428571428572e-05, |
|
"loss": 0.3984, |
|
"step": 1250 |
|
}, |
|
{ |
|
"epoch": 1.22, |
|
"learning_rate": 7.800000000000001e-05, |
|
"loss": 0.3875, |
|
"step": 1275 |
|
}, |
|
{ |
|
"epoch": 1.24, |
|
"learning_rate": 7.728571428571429e-05, |
|
"loss": 0.3925, |
|
"step": 1300 |
|
}, |
|
{ |
|
"epoch": 1.26, |
|
"learning_rate": 7.657142857142857e-05, |
|
"loss": 0.3703, |
|
"step": 1325 |
|
}, |
|
{ |
|
"epoch": 1.29, |
|
"learning_rate": 7.585714285714287e-05, |
|
"loss": 0.3898, |
|
"step": 1350 |
|
}, |
|
{ |
|
"epoch": 1.31, |
|
"learning_rate": 7.514285714285715e-05, |
|
"loss": 0.3716, |
|
"step": 1375 |
|
}, |
|
{ |
|
"epoch": 1.34, |
|
"learning_rate": 7.442857142857144e-05, |
|
"loss": 0.4206, |
|
"step": 1400 |
|
}, |
|
{ |
|
"epoch": 1.36, |
|
"learning_rate": 7.371428571428572e-05, |
|
"loss": 0.3857, |
|
"step": 1425 |
|
}, |
|
{ |
|
"epoch": 1.38, |
|
"learning_rate": 7.3e-05, |
|
"loss": 0.3527, |
|
"step": 1450 |
|
}, |
|
{ |
|
"epoch": 1.41, |
|
"learning_rate": 7.228571428571429e-05, |
|
"loss": 0.4145, |
|
"step": 1475 |
|
}, |
|
{ |
|
"epoch": 1.43, |
|
"learning_rate": 7.157142857142857e-05, |
|
"loss": 0.409, |
|
"step": 1500 |
|
}, |
|
{ |
|
"epoch": 1.46, |
|
"learning_rate": 7.085714285714285e-05, |
|
"loss": 0.3749, |
|
"step": 1525 |
|
}, |
|
{ |
|
"epoch": 1.48, |
|
"learning_rate": 7.014285714285715e-05, |
|
"loss": 0.4218, |
|
"step": 1550 |
|
}, |
|
{ |
|
"epoch": 1.5, |
|
"learning_rate": 6.942857142857143e-05, |
|
"loss": 0.3852, |
|
"step": 1575 |
|
}, |
|
{ |
|
"epoch": 1.53, |
|
"learning_rate": 6.871428571428572e-05, |
|
"loss": 0.3804, |
|
"step": 1600 |
|
}, |
|
{ |
|
"epoch": 1.55, |
|
"learning_rate": 6.800000000000001e-05, |
|
"loss": 0.4039, |
|
"step": 1625 |
|
}, |
|
{ |
|
"epoch": 1.57, |
|
"learning_rate": 6.728571428571428e-05, |
|
"loss": 0.3499, |
|
"step": 1650 |
|
}, |
|
{ |
|
"epoch": 1.6, |
|
"learning_rate": 6.657142857142857e-05, |
|
"loss": 0.386, |
|
"step": 1675 |
|
}, |
|
{ |
|
"epoch": 1.62, |
|
"learning_rate": 6.585714285714287e-05, |
|
"loss": 0.3658, |
|
"step": 1700 |
|
}, |
|
{ |
|
"epoch": 1.65, |
|
"learning_rate": 6.514285714285715e-05, |
|
"loss": 0.3653, |
|
"step": 1725 |
|
}, |
|
{ |
|
"epoch": 1.67, |
|
"learning_rate": 6.442857142857143e-05, |
|
"loss": 0.3692, |
|
"step": 1750 |
|
}, |
|
{ |
|
"epoch": 1.69, |
|
"learning_rate": 6.371428571428572e-05, |
|
"loss": 0.3852, |
|
"step": 1775 |
|
}, |
|
{ |
|
"epoch": 1.72, |
|
"learning_rate": 6.3e-05, |
|
"loss": 0.3861, |
|
"step": 1800 |
|
}, |
|
{ |
|
"epoch": 1.74, |
|
"learning_rate": 6.22857142857143e-05, |
|
"loss": 0.3845, |
|
"step": 1825 |
|
}, |
|
{ |
|
"epoch": 1.77, |
|
"learning_rate": 6.157142857142857e-05, |
|
"loss": 0.3291, |
|
"step": 1850 |
|
}, |
|
{ |
|
"epoch": 1.79, |
|
"learning_rate": 6.085714285714286e-05, |
|
"loss": 0.3558, |
|
"step": 1875 |
|
}, |
|
{ |
|
"epoch": 1.81, |
|
"learning_rate": 6.014285714285715e-05, |
|
"loss": 0.3724, |
|
"step": 1900 |
|
}, |
|
{ |
|
"epoch": 1.84, |
|
"learning_rate": 5.9428571428571434e-05, |
|
"loss": 0.3501, |
|
"step": 1925 |
|
}, |
|
{ |
|
"epoch": 1.86, |
|
"learning_rate": 5.871428571428572e-05, |
|
"loss": 0.3778, |
|
"step": 1950 |
|
}, |
|
{ |
|
"epoch": 1.88, |
|
"learning_rate": 5.8e-05, |
|
"loss": 0.362, |
|
"step": 1975 |
|
}, |
|
{ |
|
"epoch": 1.91, |
|
"learning_rate": 5.728571428571429e-05, |
|
"loss": 0.3579, |
|
"step": 2000 |
|
}, |
|
{ |
|
"epoch": 1.91, |
|
"eval_loss": 0.8363193273544312, |
|
"eval_runtime": 421.1262, |
|
"eval_samples_per_second": 3.562, |
|
"eval_steps_per_second": 0.446, |
|
"eval_wer": 51.74708818635607, |
|
"step": 2000 |
|
}, |
|
{ |
|
"epoch": 1.93, |
|
"learning_rate": 5.6571428571428574e-05, |
|
"loss": 0.363, |
|
"step": 2025 |
|
}, |
|
{ |
|
"epoch": 1.96, |
|
"learning_rate": 5.585714285714286e-05, |
|
"loss": 0.3506, |
|
"step": 2050 |
|
}, |
|
{ |
|
"epoch": 1.98, |
|
"learning_rate": 5.514285714285714e-05, |
|
"loss": 0.3653, |
|
"step": 2075 |
|
}, |
|
{ |
|
"epoch": 2.0, |
|
"learning_rate": 5.442857142857143e-05, |
|
"loss": 0.3179, |
|
"step": 2100 |
|
}, |
|
{ |
|
"epoch": 2.03, |
|
"learning_rate": 5.3714285714285714e-05, |
|
"loss": 0.1418, |
|
"step": 2125 |
|
}, |
|
{ |
|
"epoch": 2.05, |
|
"learning_rate": 5.300000000000001e-05, |
|
"loss": 0.1365, |
|
"step": 2150 |
|
}, |
|
{ |
|
"epoch": 2.08, |
|
"learning_rate": 5.2285714285714294e-05, |
|
"loss": 0.1528, |
|
"step": 2175 |
|
}, |
|
{ |
|
"epoch": 2.1, |
|
"learning_rate": 5.157142857142857e-05, |
|
"loss": 0.1419, |
|
"step": 2200 |
|
}, |
|
{ |
|
"epoch": 2.12, |
|
"learning_rate": 5.085714285714286e-05, |
|
"loss": 0.1416, |
|
"step": 2225 |
|
}, |
|
{ |
|
"epoch": 2.15, |
|
"learning_rate": 5.014285714285715e-05, |
|
"loss": 0.1435, |
|
"step": 2250 |
|
}, |
|
{ |
|
"epoch": 2.17, |
|
"learning_rate": 4.942857142857143e-05, |
|
"loss": 0.1512, |
|
"step": 2275 |
|
}, |
|
{ |
|
"epoch": 2.19, |
|
"learning_rate": 4.8714285714285714e-05, |
|
"loss": 0.1487, |
|
"step": 2300 |
|
}, |
|
{ |
|
"epoch": 2.22, |
|
"learning_rate": 4.8e-05, |
|
"loss": 0.1473, |
|
"step": 2325 |
|
}, |
|
{ |
|
"epoch": 2.24, |
|
"learning_rate": 4.728571428571429e-05, |
|
"loss": 0.1611, |
|
"step": 2350 |
|
}, |
|
{ |
|
"epoch": 2.27, |
|
"learning_rate": 4.6571428571428575e-05, |
|
"loss": 0.1471, |
|
"step": 2375 |
|
}, |
|
{ |
|
"epoch": 2.29, |
|
"learning_rate": 4.585714285714286e-05, |
|
"loss": 0.1419, |
|
"step": 2400 |
|
}, |
|
{ |
|
"epoch": 2.31, |
|
"learning_rate": 4.514285714285714e-05, |
|
"loss": 0.1311, |
|
"step": 2425 |
|
}, |
|
{ |
|
"epoch": 2.34, |
|
"learning_rate": 4.442857142857143e-05, |
|
"loss": 0.1429, |
|
"step": 2450 |
|
}, |
|
{ |
|
"epoch": 2.36, |
|
"learning_rate": 4.371428571428572e-05, |
|
"loss": 0.1364, |
|
"step": 2475 |
|
}, |
|
{ |
|
"epoch": 2.39, |
|
"learning_rate": 4.3e-05, |
|
"loss": 0.1501, |
|
"step": 2500 |
|
}, |
|
{ |
|
"epoch": 2.41, |
|
"learning_rate": 4.228571428571429e-05, |
|
"loss": 0.1525, |
|
"step": 2525 |
|
}, |
|
{ |
|
"epoch": 2.43, |
|
"learning_rate": 4.1571428571428575e-05, |
|
"loss": 0.1492, |
|
"step": 2550 |
|
}, |
|
{ |
|
"epoch": 2.46, |
|
"learning_rate": 4.085714285714286e-05, |
|
"loss": 0.1361, |
|
"step": 2575 |
|
}, |
|
{ |
|
"epoch": 2.48, |
|
"learning_rate": 4.014285714285714e-05, |
|
"loss": 0.147, |
|
"step": 2600 |
|
}, |
|
{ |
|
"epoch": 2.5, |
|
"learning_rate": 3.942857142857143e-05, |
|
"loss": 0.1399, |
|
"step": 2625 |
|
}, |
|
{ |
|
"epoch": 2.53, |
|
"learning_rate": 3.8714285714285715e-05, |
|
"loss": 0.1448, |
|
"step": 2650 |
|
}, |
|
{ |
|
"epoch": 2.55, |
|
"learning_rate": 3.8e-05, |
|
"loss": 0.1491, |
|
"step": 2675 |
|
}, |
|
{ |
|
"epoch": 2.58, |
|
"learning_rate": 3.728571428571428e-05, |
|
"loss": 0.1468, |
|
"step": 2700 |
|
}, |
|
{ |
|
"epoch": 2.6, |
|
"learning_rate": 3.6571428571428576e-05, |
|
"loss": 0.1358, |
|
"step": 2725 |
|
}, |
|
{ |
|
"epoch": 2.62, |
|
"learning_rate": 3.585714285714286e-05, |
|
"loss": 0.1387, |
|
"step": 2750 |
|
}, |
|
{ |
|
"epoch": 2.65, |
|
"learning_rate": 3.514285714285714e-05, |
|
"loss": 0.1372, |
|
"step": 2775 |
|
}, |
|
{ |
|
"epoch": 2.67, |
|
"learning_rate": 3.442857142857143e-05, |
|
"loss": 0.1342, |
|
"step": 2800 |
|
}, |
|
{ |
|
"epoch": 2.7, |
|
"learning_rate": 3.3714285714285716e-05, |
|
"loss": 0.1337, |
|
"step": 2825 |
|
}, |
|
{ |
|
"epoch": 2.72, |
|
"learning_rate": 3.3e-05, |
|
"loss": 0.1466, |
|
"step": 2850 |
|
}, |
|
{ |
|
"epoch": 2.74, |
|
"learning_rate": 3.228571428571428e-05, |
|
"loss": 0.1329, |
|
"step": 2875 |
|
}, |
|
{ |
|
"epoch": 2.77, |
|
"learning_rate": 3.1571428571428576e-05, |
|
"loss": 0.1463, |
|
"step": 2900 |
|
}, |
|
{ |
|
"epoch": 2.79, |
|
"learning_rate": 3.0857142857142856e-05, |
|
"loss": 0.1355, |
|
"step": 2925 |
|
}, |
|
{ |
|
"epoch": 2.81, |
|
"learning_rate": 3.0142857142857146e-05, |
|
"loss": 0.1395, |
|
"step": 2950 |
|
}, |
|
{ |
|
"epoch": 2.84, |
|
"learning_rate": 2.9428571428571426e-05, |
|
"loss": 0.1328, |
|
"step": 2975 |
|
}, |
|
{ |
|
"epoch": 2.86, |
|
"learning_rate": 2.8714285714285716e-05, |
|
"loss": 0.1388, |
|
"step": 3000 |
|
}, |
|
{ |
|
"epoch": 2.86, |
|
"eval_loss": 0.7765731811523438, |
|
"eval_runtime": 353.8453, |
|
"eval_samples_per_second": 4.239, |
|
"eval_steps_per_second": 0.531, |
|
"eval_wer": 43.64254021075985, |
|
"step": 3000 |
|
} |
|
], |
|
"max_steps": 4000, |
|
"num_train_epochs": 4, |
|
"total_flos": 1.18136182136832e+18, |
|
"trial_name": null, |
|
"trial_params": null |
|
} |
|
|