|
{ |
|
"best_metric": 41.3546866333888, |
|
"best_model_checkpoint": "./whisper-tiny-it-4/checkpoint-4000", |
|
"epoch": 3.816793893129771, |
|
"global_step": 4000, |
|
"is_hyper_param_search": false, |
|
"is_local_process_zero": true, |
|
"is_world_process_zero": true, |
|
"log_history": [ |
|
{ |
|
"epoch": 0.02, |
|
"learning_rate": 2.0000000000000003e-06, |
|
"loss": 3.5722, |
|
"step": 25 |
|
}, |
|
{ |
|
"epoch": 0.05, |
|
"learning_rate": 4.5e-06, |
|
"loss": 2.2272, |
|
"step": 50 |
|
}, |
|
{ |
|
"epoch": 0.07, |
|
"learning_rate": 7.000000000000001e-06, |
|
"loss": 1.1115, |
|
"step": 75 |
|
}, |
|
{ |
|
"epoch": 0.1, |
|
"learning_rate": 9.5e-06, |
|
"loss": 0.8798, |
|
"step": 100 |
|
}, |
|
{ |
|
"epoch": 0.12, |
|
"learning_rate": 1.2e-05, |
|
"loss": 0.7707, |
|
"step": 125 |
|
}, |
|
{ |
|
"epoch": 0.14, |
|
"learning_rate": 1.45e-05, |
|
"loss": 0.754, |
|
"step": 150 |
|
}, |
|
{ |
|
"epoch": 0.17, |
|
"learning_rate": 1.7000000000000003e-05, |
|
"loss": 0.7692, |
|
"step": 175 |
|
}, |
|
{ |
|
"epoch": 0.19, |
|
"learning_rate": 1.9500000000000003e-05, |
|
"loss": 0.7261, |
|
"step": 200 |
|
}, |
|
{ |
|
"epoch": 0.21, |
|
"learning_rate": 2.2000000000000003e-05, |
|
"loss": 0.661, |
|
"step": 225 |
|
}, |
|
{ |
|
"epoch": 0.24, |
|
"learning_rate": 2.45e-05, |
|
"loss": 0.6989, |
|
"step": 250 |
|
}, |
|
{ |
|
"epoch": 0.26, |
|
"learning_rate": 2.7000000000000002e-05, |
|
"loss": 0.6835, |
|
"step": 275 |
|
}, |
|
{ |
|
"epoch": 0.29, |
|
"learning_rate": 2.95e-05, |
|
"loss": 0.6839, |
|
"step": 300 |
|
}, |
|
{ |
|
"epoch": 0.31, |
|
"learning_rate": 3.2000000000000005e-05, |
|
"loss": 0.6934, |
|
"step": 325 |
|
}, |
|
{ |
|
"epoch": 0.33, |
|
"learning_rate": 3.45e-05, |
|
"loss": 0.6656, |
|
"step": 350 |
|
}, |
|
{ |
|
"epoch": 0.36, |
|
"learning_rate": 3.7e-05, |
|
"loss": 0.7053, |
|
"step": 375 |
|
}, |
|
{ |
|
"epoch": 0.38, |
|
"learning_rate": 3.9500000000000005e-05, |
|
"loss": 0.6652, |
|
"step": 400 |
|
}, |
|
{ |
|
"epoch": 0.41, |
|
"learning_rate": 4.2e-05, |
|
"loss": 0.6747, |
|
"step": 425 |
|
}, |
|
{ |
|
"epoch": 0.43, |
|
"learning_rate": 4.4500000000000004e-05, |
|
"loss": 0.6654, |
|
"step": 450 |
|
}, |
|
{ |
|
"epoch": 0.45, |
|
"learning_rate": 4.7e-05, |
|
"loss": 0.6408, |
|
"step": 475 |
|
}, |
|
{ |
|
"epoch": 0.48, |
|
"learning_rate": 4.9500000000000004e-05, |
|
"loss": 0.6722, |
|
"step": 500 |
|
}, |
|
{ |
|
"epoch": 0.5, |
|
"learning_rate": 4.971428571428572e-05, |
|
"loss": 0.6924, |
|
"step": 525 |
|
}, |
|
{ |
|
"epoch": 0.52, |
|
"learning_rate": 4.935714285714286e-05, |
|
"loss": 0.6558, |
|
"step": 550 |
|
}, |
|
{ |
|
"epoch": 0.55, |
|
"learning_rate": 4.9e-05, |
|
"loss": 0.6947, |
|
"step": 575 |
|
}, |
|
{ |
|
"epoch": 0.57, |
|
"learning_rate": 4.8642857142857145e-05, |
|
"loss": 0.6687, |
|
"step": 600 |
|
}, |
|
{ |
|
"epoch": 0.6, |
|
"learning_rate": 4.828571428571429e-05, |
|
"loss": 0.6956, |
|
"step": 625 |
|
}, |
|
{ |
|
"epoch": 0.62, |
|
"learning_rate": 4.7928571428571425e-05, |
|
"loss": 0.6678, |
|
"step": 650 |
|
}, |
|
{ |
|
"epoch": 0.64, |
|
"learning_rate": 4.757142857142857e-05, |
|
"loss": 0.6195, |
|
"step": 675 |
|
}, |
|
{ |
|
"epoch": 0.67, |
|
"learning_rate": 4.721428571428572e-05, |
|
"loss": 0.652, |
|
"step": 700 |
|
}, |
|
{ |
|
"epoch": 0.69, |
|
"learning_rate": 4.685714285714286e-05, |
|
"loss": 0.6484, |
|
"step": 725 |
|
}, |
|
{ |
|
"epoch": 0.72, |
|
"learning_rate": 4.6500000000000005e-05, |
|
"loss": 0.6104, |
|
"step": 750 |
|
}, |
|
{ |
|
"epoch": 0.74, |
|
"learning_rate": 4.6142857142857145e-05, |
|
"loss": 0.6369, |
|
"step": 775 |
|
}, |
|
{ |
|
"epoch": 0.76, |
|
"learning_rate": 4.5785714285714285e-05, |
|
"loss": 0.586, |
|
"step": 800 |
|
}, |
|
{ |
|
"epoch": 0.79, |
|
"learning_rate": 4.542857142857143e-05, |
|
"loss": 0.6088, |
|
"step": 825 |
|
}, |
|
{ |
|
"epoch": 0.81, |
|
"learning_rate": 4.507142857142858e-05, |
|
"loss": 0.6312, |
|
"step": 850 |
|
}, |
|
{ |
|
"epoch": 0.83, |
|
"learning_rate": 4.471428571428571e-05, |
|
"loss": 0.6064, |
|
"step": 875 |
|
}, |
|
{ |
|
"epoch": 0.86, |
|
"learning_rate": 4.435714285714286e-05, |
|
"loss": 0.6264, |
|
"step": 900 |
|
}, |
|
{ |
|
"epoch": 0.88, |
|
"learning_rate": 4.4000000000000006e-05, |
|
"loss": 0.6792, |
|
"step": 925 |
|
}, |
|
{ |
|
"epoch": 0.91, |
|
"learning_rate": 4.3642857142857146e-05, |
|
"loss": 0.5809, |
|
"step": 950 |
|
}, |
|
{ |
|
"epoch": 0.93, |
|
"learning_rate": 4.328571428571429e-05, |
|
"loss": 0.6006, |
|
"step": 975 |
|
}, |
|
{ |
|
"epoch": 0.95, |
|
"learning_rate": 4.292857142857143e-05, |
|
"loss": 0.5919, |
|
"step": 1000 |
|
}, |
|
{ |
|
"epoch": 0.95, |
|
"eval_loss": 0.804903507232666, |
|
"eval_runtime": 401.2067, |
|
"eval_samples_per_second": 3.739, |
|
"eval_steps_per_second": 0.469, |
|
"eval_wer": 56.48225180255131, |
|
"step": 1000 |
|
}, |
|
{ |
|
"epoch": 0.98, |
|
"learning_rate": 4.257142857142857e-05, |
|
"loss": 0.6182, |
|
"step": 1025 |
|
}, |
|
{ |
|
"epoch": 1.0, |
|
"learning_rate": 4.221428571428572e-05, |
|
"loss": 0.5634, |
|
"step": 1050 |
|
}, |
|
{ |
|
"epoch": 1.03, |
|
"learning_rate": 4.185714285714286e-05, |
|
"loss": 0.3226, |
|
"step": 1075 |
|
}, |
|
{ |
|
"epoch": 1.05, |
|
"learning_rate": 4.15e-05, |
|
"loss": 0.3305, |
|
"step": 1100 |
|
}, |
|
{ |
|
"epoch": 1.07, |
|
"learning_rate": 4.1142857142857146e-05, |
|
"loss": 0.3342, |
|
"step": 1125 |
|
}, |
|
{ |
|
"epoch": 1.1, |
|
"learning_rate": 4.0785714285714286e-05, |
|
"loss": 0.3565, |
|
"step": 1150 |
|
}, |
|
{ |
|
"epoch": 1.12, |
|
"learning_rate": 4.042857142857143e-05, |
|
"loss": 0.3389, |
|
"step": 1175 |
|
}, |
|
{ |
|
"epoch": 1.15, |
|
"learning_rate": 4.007142857142857e-05, |
|
"loss": 0.3322, |
|
"step": 1200 |
|
}, |
|
{ |
|
"epoch": 1.17, |
|
"learning_rate": 3.971428571428571e-05, |
|
"loss": 0.3462, |
|
"step": 1225 |
|
}, |
|
{ |
|
"epoch": 1.19, |
|
"learning_rate": 3.935714285714286e-05, |
|
"loss": 0.3442, |
|
"step": 1250 |
|
}, |
|
{ |
|
"epoch": 1.22, |
|
"learning_rate": 3.9000000000000006e-05, |
|
"loss": 0.3433, |
|
"step": 1275 |
|
}, |
|
{ |
|
"epoch": 1.24, |
|
"learning_rate": 3.8642857142857146e-05, |
|
"loss": 0.3398, |
|
"step": 1300 |
|
}, |
|
{ |
|
"epoch": 1.26, |
|
"learning_rate": 3.8285714285714286e-05, |
|
"loss": 0.32, |
|
"step": 1325 |
|
}, |
|
{ |
|
"epoch": 1.29, |
|
"learning_rate": 3.792857142857143e-05, |
|
"loss": 0.3482, |
|
"step": 1350 |
|
}, |
|
{ |
|
"epoch": 1.31, |
|
"learning_rate": 3.757142857142857e-05, |
|
"loss": 0.3189, |
|
"step": 1375 |
|
}, |
|
{ |
|
"epoch": 1.34, |
|
"learning_rate": 3.721428571428572e-05, |
|
"loss": 0.3682, |
|
"step": 1400 |
|
}, |
|
{ |
|
"epoch": 1.36, |
|
"learning_rate": 3.685714285714286e-05, |
|
"loss": 0.3379, |
|
"step": 1425 |
|
}, |
|
{ |
|
"epoch": 1.38, |
|
"learning_rate": 3.65e-05, |
|
"loss": 0.3116, |
|
"step": 1450 |
|
}, |
|
{ |
|
"epoch": 1.41, |
|
"learning_rate": 3.6142857142857146e-05, |
|
"loss": 0.354, |
|
"step": 1475 |
|
}, |
|
{ |
|
"epoch": 1.43, |
|
"learning_rate": 3.5785714285714286e-05, |
|
"loss": 0.3559, |
|
"step": 1500 |
|
}, |
|
{ |
|
"epoch": 1.46, |
|
"learning_rate": 3.5428571428571426e-05, |
|
"loss": 0.3395, |
|
"step": 1525 |
|
}, |
|
{ |
|
"epoch": 1.48, |
|
"learning_rate": 3.507142857142857e-05, |
|
"loss": 0.3693, |
|
"step": 1550 |
|
}, |
|
{ |
|
"epoch": 1.5, |
|
"learning_rate": 3.471428571428571e-05, |
|
"loss": 0.3307, |
|
"step": 1575 |
|
}, |
|
{ |
|
"epoch": 1.53, |
|
"learning_rate": 3.435714285714286e-05, |
|
"loss": 0.3259, |
|
"step": 1600 |
|
}, |
|
{ |
|
"epoch": 1.55, |
|
"learning_rate": 3.4000000000000007e-05, |
|
"loss": 0.3576, |
|
"step": 1625 |
|
}, |
|
{ |
|
"epoch": 1.57, |
|
"learning_rate": 3.364285714285714e-05, |
|
"loss": 0.3083, |
|
"step": 1650 |
|
}, |
|
{ |
|
"epoch": 1.6, |
|
"learning_rate": 3.3285714285714286e-05, |
|
"loss": 0.3413, |
|
"step": 1675 |
|
}, |
|
{ |
|
"epoch": 1.62, |
|
"learning_rate": 3.292857142857143e-05, |
|
"loss": 0.3178, |
|
"step": 1700 |
|
}, |
|
{ |
|
"epoch": 1.65, |
|
"learning_rate": 3.257142857142857e-05, |
|
"loss": 0.3331, |
|
"step": 1725 |
|
}, |
|
{ |
|
"epoch": 1.67, |
|
"learning_rate": 3.221428571428571e-05, |
|
"loss": 0.3208, |
|
"step": 1750 |
|
}, |
|
{ |
|
"epoch": 1.69, |
|
"learning_rate": 3.185714285714286e-05, |
|
"loss": 0.3275, |
|
"step": 1775 |
|
}, |
|
{ |
|
"epoch": 1.72, |
|
"learning_rate": 3.15e-05, |
|
"loss": 0.3368, |
|
"step": 1800 |
|
}, |
|
{ |
|
"epoch": 1.74, |
|
"learning_rate": 3.114285714285715e-05, |
|
"loss": 0.3335, |
|
"step": 1825 |
|
}, |
|
{ |
|
"epoch": 1.77, |
|
"learning_rate": 3.078571428571429e-05, |
|
"loss": 0.2934, |
|
"step": 1850 |
|
}, |
|
{ |
|
"epoch": 1.79, |
|
"learning_rate": 3.042857142857143e-05, |
|
"loss": 0.3081, |
|
"step": 1875 |
|
}, |
|
{ |
|
"epoch": 1.81, |
|
"learning_rate": 3.0071428571428573e-05, |
|
"loss": 0.3298, |
|
"step": 1900 |
|
}, |
|
{ |
|
"epoch": 1.84, |
|
"learning_rate": 2.9714285714285717e-05, |
|
"loss": 0.3052, |
|
"step": 1925 |
|
}, |
|
{ |
|
"epoch": 1.86, |
|
"learning_rate": 2.935714285714286e-05, |
|
"loss": 0.3281, |
|
"step": 1950 |
|
}, |
|
{ |
|
"epoch": 1.88, |
|
"learning_rate": 2.9e-05, |
|
"loss": 0.3194, |
|
"step": 1975 |
|
}, |
|
{ |
|
"epoch": 1.91, |
|
"learning_rate": 2.8642857142857144e-05, |
|
"loss": 0.3181, |
|
"step": 2000 |
|
}, |
|
{ |
|
"epoch": 1.91, |
|
"eval_loss": 0.7393131256103516, |
|
"eval_runtime": 356.4776, |
|
"eval_samples_per_second": 4.208, |
|
"eval_steps_per_second": 0.527, |
|
"eval_wer": 44.81419855795895, |
|
"step": 2000 |
|
}, |
|
{ |
|
"epoch": 1.93, |
|
"learning_rate": 2.8285714285714287e-05, |
|
"loss": 0.3146, |
|
"step": 2025 |
|
}, |
|
{ |
|
"epoch": 1.96, |
|
"learning_rate": 2.792857142857143e-05, |
|
"loss": 0.3081, |
|
"step": 2050 |
|
}, |
|
{ |
|
"epoch": 1.98, |
|
"learning_rate": 2.757142857142857e-05, |
|
"loss": 0.3204, |
|
"step": 2075 |
|
}, |
|
{ |
|
"epoch": 2.0, |
|
"learning_rate": 2.7214285714285714e-05, |
|
"loss": 0.2845, |
|
"step": 2100 |
|
}, |
|
{ |
|
"epoch": 2.03, |
|
"learning_rate": 2.6857142857142857e-05, |
|
"loss": 0.1397, |
|
"step": 2125 |
|
}, |
|
{ |
|
"epoch": 2.05, |
|
"learning_rate": 2.6500000000000004e-05, |
|
"loss": 0.1343, |
|
"step": 2150 |
|
}, |
|
{ |
|
"epoch": 2.08, |
|
"learning_rate": 2.6142857142857147e-05, |
|
"loss": 0.146, |
|
"step": 2175 |
|
}, |
|
{ |
|
"epoch": 2.1, |
|
"learning_rate": 2.5785714285714284e-05, |
|
"loss": 0.1424, |
|
"step": 2200 |
|
}, |
|
{ |
|
"epoch": 2.12, |
|
"learning_rate": 2.542857142857143e-05, |
|
"loss": 0.1407, |
|
"step": 2225 |
|
}, |
|
{ |
|
"epoch": 2.15, |
|
"learning_rate": 2.5071428571428574e-05, |
|
"loss": 0.14, |
|
"step": 2250 |
|
}, |
|
{ |
|
"epoch": 2.17, |
|
"learning_rate": 2.4714285714285714e-05, |
|
"loss": 0.1545, |
|
"step": 2275 |
|
}, |
|
{ |
|
"epoch": 2.19, |
|
"learning_rate": 2.4357142857142857e-05, |
|
"loss": 0.1481, |
|
"step": 2300 |
|
}, |
|
{ |
|
"epoch": 2.22, |
|
"learning_rate": 2.4e-05, |
|
"loss": 0.1475, |
|
"step": 2325 |
|
}, |
|
{ |
|
"epoch": 2.24, |
|
"learning_rate": 2.3642857142857144e-05, |
|
"loss": 0.1641, |
|
"step": 2350 |
|
}, |
|
{ |
|
"epoch": 2.27, |
|
"learning_rate": 2.3285714285714287e-05, |
|
"loss": 0.1418, |
|
"step": 2375 |
|
}, |
|
{ |
|
"epoch": 2.29, |
|
"learning_rate": 2.292857142857143e-05, |
|
"loss": 0.1476, |
|
"step": 2400 |
|
}, |
|
{ |
|
"epoch": 2.31, |
|
"learning_rate": 2.257142857142857e-05, |
|
"loss": 0.1299, |
|
"step": 2425 |
|
}, |
|
{ |
|
"epoch": 2.34, |
|
"learning_rate": 2.2214285714285714e-05, |
|
"loss": 0.1354, |
|
"step": 2450 |
|
}, |
|
{ |
|
"epoch": 2.36, |
|
"learning_rate": 2.185714285714286e-05, |
|
"loss": 0.1379, |
|
"step": 2475 |
|
}, |
|
{ |
|
"epoch": 2.39, |
|
"learning_rate": 2.15e-05, |
|
"loss": 0.1493, |
|
"step": 2500 |
|
}, |
|
{ |
|
"epoch": 2.41, |
|
"learning_rate": 2.1142857142857144e-05, |
|
"loss": 0.1474, |
|
"step": 2525 |
|
}, |
|
{ |
|
"epoch": 2.43, |
|
"learning_rate": 2.0785714285714288e-05, |
|
"loss": 0.147, |
|
"step": 2550 |
|
}, |
|
{ |
|
"epoch": 2.46, |
|
"learning_rate": 2.042857142857143e-05, |
|
"loss": 0.1495, |
|
"step": 2575 |
|
}, |
|
{ |
|
"epoch": 2.48, |
|
"learning_rate": 2.007142857142857e-05, |
|
"loss": 0.1502, |
|
"step": 2600 |
|
}, |
|
{ |
|
"epoch": 2.5, |
|
"learning_rate": 1.9714285714285714e-05, |
|
"loss": 0.147, |
|
"step": 2625 |
|
}, |
|
{ |
|
"epoch": 2.53, |
|
"learning_rate": 1.9357142857142858e-05, |
|
"loss": 0.148, |
|
"step": 2650 |
|
}, |
|
{ |
|
"epoch": 2.55, |
|
"learning_rate": 1.9e-05, |
|
"loss": 0.1493, |
|
"step": 2675 |
|
}, |
|
{ |
|
"epoch": 2.58, |
|
"learning_rate": 1.864285714285714e-05, |
|
"loss": 0.1469, |
|
"step": 2700 |
|
}, |
|
{ |
|
"epoch": 2.6, |
|
"learning_rate": 1.8285714285714288e-05, |
|
"loss": 0.1372, |
|
"step": 2725 |
|
}, |
|
{ |
|
"epoch": 2.62, |
|
"learning_rate": 1.792857142857143e-05, |
|
"loss": 0.1468, |
|
"step": 2750 |
|
}, |
|
{ |
|
"epoch": 2.65, |
|
"learning_rate": 1.757142857142857e-05, |
|
"loss": 0.1418, |
|
"step": 2775 |
|
}, |
|
{ |
|
"epoch": 2.67, |
|
"learning_rate": 1.7214285714285715e-05, |
|
"loss": 0.1389, |
|
"step": 2800 |
|
}, |
|
{ |
|
"epoch": 2.7, |
|
"learning_rate": 1.6857142857142858e-05, |
|
"loss": 0.1347, |
|
"step": 2825 |
|
}, |
|
{ |
|
"epoch": 2.72, |
|
"learning_rate": 1.65e-05, |
|
"loss": 0.1508, |
|
"step": 2850 |
|
}, |
|
{ |
|
"epoch": 2.74, |
|
"learning_rate": 1.614285714285714e-05, |
|
"loss": 0.1347, |
|
"step": 2875 |
|
}, |
|
{ |
|
"epoch": 2.77, |
|
"learning_rate": 1.5785714285714288e-05, |
|
"loss": 0.143, |
|
"step": 2900 |
|
}, |
|
{ |
|
"epoch": 2.79, |
|
"learning_rate": 1.5428571428571428e-05, |
|
"loss": 0.1433, |
|
"step": 2925 |
|
}, |
|
{ |
|
"epoch": 2.81, |
|
"learning_rate": 1.5071428571428573e-05, |
|
"loss": 0.1532, |
|
"step": 2950 |
|
}, |
|
{ |
|
"epoch": 2.84, |
|
"learning_rate": 1.4714285714285713e-05, |
|
"loss": 0.1472, |
|
"step": 2975 |
|
}, |
|
{ |
|
"epoch": 2.86, |
|
"learning_rate": 1.4357142857142858e-05, |
|
"loss": 0.1417, |
|
"step": 3000 |
|
}, |
|
{ |
|
"epoch": 2.86, |
|
"eval_loss": 0.7067434787750244, |
|
"eval_runtime": 360.1108, |
|
"eval_samples_per_second": 4.165, |
|
"eval_steps_per_second": 0.522, |
|
"eval_wer": 42.748197448696615, |
|
"step": 3000 |
|
}, |
|
{ |
|
"epoch": 2.89, |
|
"learning_rate": 1.4000000000000001e-05, |
|
"loss": 0.1477, |
|
"step": 3025 |
|
}, |
|
{ |
|
"epoch": 2.91, |
|
"learning_rate": 1.3642857142857143e-05, |
|
"loss": 0.1466, |
|
"step": 3050 |
|
}, |
|
{ |
|
"epoch": 2.93, |
|
"learning_rate": 1.3285714285714288e-05, |
|
"loss": 0.1388, |
|
"step": 3075 |
|
}, |
|
{ |
|
"epoch": 2.96, |
|
"learning_rate": 1.2928571428571428e-05, |
|
"loss": 0.1417, |
|
"step": 3100 |
|
}, |
|
{ |
|
"epoch": 2.98, |
|
"learning_rate": 1.2571428571428573e-05, |
|
"loss": 0.1316, |
|
"step": 3125 |
|
}, |
|
{ |
|
"epoch": 3.01, |
|
"learning_rate": 1.2214285714285715e-05, |
|
"loss": 0.1185, |
|
"step": 3150 |
|
}, |
|
{ |
|
"epoch": 3.03, |
|
"learning_rate": 1.1857142857142858e-05, |
|
"loss": 0.061, |
|
"step": 3175 |
|
}, |
|
{ |
|
"epoch": 3.05, |
|
"learning_rate": 1.1500000000000002e-05, |
|
"loss": 0.0632, |
|
"step": 3200 |
|
}, |
|
{ |
|
"epoch": 3.08, |
|
"learning_rate": 1.1142857142857143e-05, |
|
"loss": 0.0589, |
|
"step": 3225 |
|
}, |
|
{ |
|
"epoch": 3.1, |
|
"learning_rate": 1.0785714285714287e-05, |
|
"loss": 0.0645, |
|
"step": 3250 |
|
}, |
|
{ |
|
"epoch": 3.12, |
|
"learning_rate": 1.0428571428571428e-05, |
|
"loss": 0.0623, |
|
"step": 3275 |
|
}, |
|
{ |
|
"epoch": 3.15, |
|
"learning_rate": 1.0071428571428572e-05, |
|
"loss": 0.064, |
|
"step": 3300 |
|
}, |
|
{ |
|
"epoch": 3.17, |
|
"learning_rate": 9.714285714285715e-06, |
|
"loss": 0.0623, |
|
"step": 3325 |
|
}, |
|
{ |
|
"epoch": 3.2, |
|
"learning_rate": 9.357142857142857e-06, |
|
"loss": 0.0576, |
|
"step": 3350 |
|
}, |
|
{ |
|
"epoch": 3.22, |
|
"learning_rate": 9e-06, |
|
"loss": 0.0654, |
|
"step": 3375 |
|
}, |
|
{ |
|
"epoch": 3.24, |
|
"learning_rate": 8.642857142857144e-06, |
|
"loss": 0.0652, |
|
"step": 3400 |
|
}, |
|
{ |
|
"epoch": 3.27, |
|
"learning_rate": 8.285714285714285e-06, |
|
"loss": 0.0575, |
|
"step": 3425 |
|
}, |
|
{ |
|
"epoch": 3.29, |
|
"learning_rate": 7.928571428571429e-06, |
|
"loss": 0.062, |
|
"step": 3450 |
|
}, |
|
{ |
|
"epoch": 3.32, |
|
"learning_rate": 7.571428571428572e-06, |
|
"loss": 0.0594, |
|
"step": 3475 |
|
}, |
|
{ |
|
"epoch": 3.34, |
|
"learning_rate": 7.214285714285715e-06, |
|
"loss": 0.063, |
|
"step": 3500 |
|
}, |
|
{ |
|
"epoch": 3.36, |
|
"learning_rate": 6.857142857142858e-06, |
|
"loss": 0.0638, |
|
"step": 3525 |
|
}, |
|
{ |
|
"epoch": 3.39, |
|
"learning_rate": 6.5000000000000004e-06, |
|
"loss": 0.0612, |
|
"step": 3550 |
|
}, |
|
{ |
|
"epoch": 3.41, |
|
"learning_rate": 6.142857142857143e-06, |
|
"loss": 0.0644, |
|
"step": 3575 |
|
}, |
|
{ |
|
"epoch": 3.44, |
|
"learning_rate": 5.785714285714286e-06, |
|
"loss": 0.0592, |
|
"step": 3600 |
|
}, |
|
{ |
|
"epoch": 3.46, |
|
"learning_rate": 5.428571428571429e-06, |
|
"loss": 0.06, |
|
"step": 3625 |
|
}, |
|
{ |
|
"epoch": 3.48, |
|
"learning_rate": 5.071428571428571e-06, |
|
"loss": 0.0552, |
|
"step": 3650 |
|
}, |
|
{ |
|
"epoch": 3.51, |
|
"learning_rate": 4.714285714285715e-06, |
|
"loss": 0.0607, |
|
"step": 3675 |
|
}, |
|
{ |
|
"epoch": 3.53, |
|
"learning_rate": 4.357142857142857e-06, |
|
"loss": 0.0609, |
|
"step": 3700 |
|
}, |
|
{ |
|
"epoch": 3.55, |
|
"learning_rate": 4.000000000000001e-06, |
|
"loss": 0.0638, |
|
"step": 3725 |
|
}, |
|
{ |
|
"epoch": 3.58, |
|
"learning_rate": 3.642857142857143e-06, |
|
"loss": 0.0594, |
|
"step": 3750 |
|
}, |
|
{ |
|
"epoch": 3.6, |
|
"learning_rate": 3.2857142857142857e-06, |
|
"loss": 0.0666, |
|
"step": 3775 |
|
}, |
|
{ |
|
"epoch": 3.63, |
|
"learning_rate": 2.9285714285714287e-06, |
|
"loss": 0.0539, |
|
"step": 3800 |
|
}, |
|
{ |
|
"epoch": 3.65, |
|
"learning_rate": 2.5714285714285716e-06, |
|
"loss": 0.0612, |
|
"step": 3825 |
|
}, |
|
{ |
|
"epoch": 3.67, |
|
"learning_rate": 2.214285714285714e-06, |
|
"loss": 0.0581, |
|
"step": 3850 |
|
}, |
|
{ |
|
"epoch": 3.7, |
|
"learning_rate": 1.8571428571428573e-06, |
|
"loss": 0.0649, |
|
"step": 3875 |
|
}, |
|
{ |
|
"epoch": 3.72, |
|
"learning_rate": 1.5e-06, |
|
"loss": 0.0571, |
|
"step": 3900 |
|
}, |
|
{ |
|
"epoch": 3.75, |
|
"learning_rate": 1.142857142857143e-06, |
|
"loss": 0.0582, |
|
"step": 3925 |
|
}, |
|
{ |
|
"epoch": 3.77, |
|
"learning_rate": 7.857142857142858e-07, |
|
"loss": 0.0597, |
|
"step": 3950 |
|
}, |
|
{ |
|
"epoch": 3.79, |
|
"learning_rate": 4.285714285714286e-07, |
|
"loss": 0.0547, |
|
"step": 3975 |
|
}, |
|
{ |
|
"epoch": 3.82, |
|
"learning_rate": 7.142857142857144e-08, |
|
"loss": 0.0627, |
|
"step": 4000 |
|
}, |
|
{ |
|
"epoch": 3.82, |
|
"eval_loss": 0.7125627994537354, |
|
"eval_runtime": 353.3354, |
|
"eval_samples_per_second": 4.245, |
|
"eval_steps_per_second": 0.532, |
|
"eval_wer": 41.3546866333888, |
|
"step": 4000 |
|
} |
|
], |
|
"max_steps": 4000, |
|
"num_train_epochs": 4, |
|
"total_flos": 1.57509165109248e+18, |
|
"trial_name": null, |
|
"trial_params": null |
|
} |
|
|