| { | |
| "best_metric": 9.50037147102526, | |
| "best_model_checkpoint": "./data/finetuningRuns/whisper-lg-el-intlv-xs-2/checkpoint-4000", | |
| "epoch": 12.437810945273633, | |
| "global_step": 5000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.5399999999999998e-07, | |
| "loss": 3.2988, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 3.29e-07, | |
| "loss": 2.5807, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 5.04e-07, | |
| "loss": 1.8704, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 6.79e-07, | |
| "loss": 1.4293, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 8.54e-07, | |
| "loss": 1.1838, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.029e-06, | |
| "loss": 0.9288, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.1970000000000001e-06, | |
| "loss": 0.7441, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.372e-06, | |
| "loss": 0.6405, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.547e-06, | |
| "loss": 0.5809, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.722e-06, | |
| "loss": 0.499, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.897e-06, | |
| "loss": 0.4434, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 2.072e-06, | |
| "loss": 0.3797, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 2.247e-06, | |
| "loss": 0.3427, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 2.422e-06, | |
| "loss": 0.3135, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.597e-06, | |
| "loss": 0.2714, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 2.772e-06, | |
| "loss": 0.2658, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 2.947e-06, | |
| "loss": 0.2422, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 3.122e-06, | |
| "loss": 0.2144, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 3.297e-06, | |
| "loss": 0.21, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 3.472e-06, | |
| "loss": 0.1836, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 3.4836666666666664e-06, | |
| "loss": 0.1731, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 3.464222222222222e-06, | |
| "loss": 0.1698, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 3.4447777777777776e-06, | |
| "loss": 0.1731, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 3.4253333333333335e-06, | |
| "loss": 0.1657, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 3.405888888888889e-06, | |
| "loss": 0.1458, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 3.3864444444444443e-06, | |
| "loss": 0.1322, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 3.3669999999999997e-06, | |
| "loss": 0.1321, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 3.3475555555555555e-06, | |
| "loss": 0.1396, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 3.328111111111111e-06, | |
| "loss": 0.1248, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 3.3086666666666668e-06, | |
| "loss": 0.1125, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 3.289222222222222e-06, | |
| "loss": 0.1295, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 3.2697777777777776e-06, | |
| "loss": 0.1148, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 3.250333333333333e-06, | |
| "loss": 0.1057, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 3.230888888888889e-06, | |
| "loss": 0.0915, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 3.2114444444444443e-06, | |
| "loss": 0.0857, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 3.192e-06, | |
| "loss": 0.0933, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 3.1725555555555555e-06, | |
| "loss": 0.0875, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 3.153111111111111e-06, | |
| "loss": 0.0926, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 3.1336666666666668e-06, | |
| "loss": 0.0842, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "learning_rate": 3.114222222222222e-06, | |
| "loss": 0.0813, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "eval_loss": 0.21469815075397491, | |
| "eval_runtime": 1983.3845, | |
| "eval_samples_per_second": 0.855, | |
| "eval_steps_per_second": 0.107, | |
| "eval_wer": 10.828380386329867, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 3.094777777777778e-06, | |
| "loss": 0.0804, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 2.61, | |
| "learning_rate": 3.0753333333333334e-06, | |
| "loss": 0.0861, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 3.055888888888889e-06, | |
| "loss": 0.0801, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 2.74, | |
| "learning_rate": 3.0364444444444442e-06, | |
| "loss": 0.0798, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "learning_rate": 3.017e-06, | |
| "loss": 0.0758, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 2.86, | |
| "learning_rate": 2.9975555555555555e-06, | |
| "loss": 0.0764, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "learning_rate": 2.9781111111111113e-06, | |
| "loss": 0.0763, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "learning_rate": 2.9586666666666667e-06, | |
| "loss": 0.0738, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 2.939222222222222e-06, | |
| "loss": 0.063, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "learning_rate": 2.9197777777777776e-06, | |
| "loss": 0.0612, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 3.17, | |
| "learning_rate": 2.9003333333333334e-06, | |
| "loss": 0.0639, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 3.23, | |
| "learning_rate": 2.880888888888889e-06, | |
| "loss": 0.0564, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 3.3, | |
| "learning_rate": 2.8614444444444447e-06, | |
| "loss": 0.0576, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 3.36, | |
| "learning_rate": 2.842e-06, | |
| "loss": 0.0591, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 3.42, | |
| "learning_rate": 2.8225555555555555e-06, | |
| "loss": 0.0541, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 3.48, | |
| "learning_rate": 2.803111111111111e-06, | |
| "loss": 0.057, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 3.54, | |
| "learning_rate": 2.7836666666666667e-06, | |
| "loss": 0.0619, | |
| "step": 1425 | |
| }, | |
| { | |
| "epoch": 3.61, | |
| "learning_rate": 2.764222222222222e-06, | |
| "loss": 0.0592, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 3.67, | |
| "learning_rate": 2.744777777777778e-06, | |
| "loss": 0.0605, | |
| "step": 1475 | |
| }, | |
| { | |
| "epoch": 3.73, | |
| "learning_rate": 2.725333333333333e-06, | |
| "loss": 0.0551, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 3.79, | |
| "learning_rate": 2.705888888888889e-06, | |
| "loss": 0.0477, | |
| "step": 1525 | |
| }, | |
| { | |
| "epoch": 3.86, | |
| "learning_rate": 2.686444444444444e-06, | |
| "loss": 0.0497, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 3.92, | |
| "learning_rate": 2.667e-06, | |
| "loss": 0.0589, | |
| "step": 1575 | |
| }, | |
| { | |
| "epoch": 3.98, | |
| "learning_rate": 2.6475555555555555e-06, | |
| "loss": 0.0572, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 4.04, | |
| "learning_rate": 2.6281111111111113e-06, | |
| "loss": 0.0474, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 4.1, | |
| "learning_rate": 2.6086666666666663e-06, | |
| "loss": 0.0432, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 4.17, | |
| "learning_rate": 2.589222222222222e-06, | |
| "loss": 0.0377, | |
| "step": 1675 | |
| }, | |
| { | |
| "epoch": 4.23, | |
| "learning_rate": 2.5697777777777775e-06, | |
| "loss": 0.0449, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 4.29, | |
| "learning_rate": 2.5503333333333334e-06, | |
| "loss": 0.0413, | |
| "step": 1725 | |
| }, | |
| { | |
| "epoch": 4.35, | |
| "learning_rate": 2.5308888888888888e-06, | |
| "loss": 0.0415, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 4.42, | |
| "learning_rate": 2.511444444444444e-06, | |
| "loss": 0.0435, | |
| "step": 1775 | |
| }, | |
| { | |
| "epoch": 4.48, | |
| "learning_rate": 2.492e-06, | |
| "loss": 0.0449, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 4.54, | |
| "learning_rate": 2.4725555555555554e-06, | |
| "loss": 0.0435, | |
| "step": 1825 | |
| }, | |
| { | |
| "epoch": 4.6, | |
| "learning_rate": 2.4531111111111113e-06, | |
| "loss": 0.0416, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 4.66, | |
| "learning_rate": 2.4336666666666667e-06, | |
| "loss": 0.045, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 4.73, | |
| "learning_rate": 2.4142222222222225e-06, | |
| "loss": 0.0444, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 4.79, | |
| "learning_rate": 2.3947777777777775e-06, | |
| "loss": 0.042, | |
| "step": 1925 | |
| }, | |
| { | |
| "epoch": 4.85, | |
| "learning_rate": 2.3753333333333333e-06, | |
| "loss": 0.0368, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 4.91, | |
| "learning_rate": 2.3558888888888888e-06, | |
| "loss": 0.0392, | |
| "step": 1975 | |
| }, | |
| { | |
| "epoch": 4.98, | |
| "learning_rate": 2.3364444444444446e-06, | |
| "loss": 0.0379, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 4.98, | |
| "eval_loss": 0.24391739070415497, | |
| "eval_runtime": 1980.3679, | |
| "eval_samples_per_second": 0.856, | |
| "eval_steps_per_second": 0.107, | |
| "eval_wer": 10.0111441307578, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 5.04, | |
| "learning_rate": 2.317e-06, | |
| "loss": 0.0362, | |
| "step": 2025 | |
| }, | |
| { | |
| "epoch": 5.1, | |
| "learning_rate": 2.297555555555556e-06, | |
| "loss": 0.0307, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 5.16, | |
| "learning_rate": 2.278111111111111e-06, | |
| "loss": 0.0299, | |
| "step": 2075 | |
| }, | |
| { | |
| "epoch": 5.22, | |
| "learning_rate": 2.2586666666666667e-06, | |
| "loss": 0.0293, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 5.29, | |
| "learning_rate": 2.239222222222222e-06, | |
| "loss": 0.034, | |
| "step": 2125 | |
| }, | |
| { | |
| "epoch": 5.35, | |
| "learning_rate": 2.219777777777778e-06, | |
| "loss": 0.0315, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 5.41, | |
| "learning_rate": 2.2003333333333333e-06, | |
| "loss": 0.0337, | |
| "step": 2175 | |
| }, | |
| { | |
| "epoch": 5.47, | |
| "learning_rate": 2.1808888888888887e-06, | |
| "loss": 0.0338, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 5.53, | |
| "learning_rate": 2.161444444444444e-06, | |
| "loss": 0.0333, | |
| "step": 2225 | |
| }, | |
| { | |
| "epoch": 5.6, | |
| "learning_rate": 2.142e-06, | |
| "loss": 0.0319, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 5.66, | |
| "learning_rate": 2.1225555555555554e-06, | |
| "loss": 0.0303, | |
| "step": 2275 | |
| }, | |
| { | |
| "epoch": 5.72, | |
| "learning_rate": 2.1031111111111112e-06, | |
| "loss": 0.0322, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 5.78, | |
| "learning_rate": 2.0836666666666667e-06, | |
| "loss": 0.0309, | |
| "step": 2325 | |
| }, | |
| { | |
| "epoch": 5.85, | |
| "learning_rate": 2.064222222222222e-06, | |
| "loss": 0.0315, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 5.91, | |
| "learning_rate": 2.0447777777777775e-06, | |
| "loss": 0.035, | |
| "step": 2375 | |
| }, | |
| { | |
| "epoch": 5.97, | |
| "learning_rate": 2.0253333333333333e-06, | |
| "loss": 0.0256, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 6.03, | |
| "learning_rate": 2.0058888888888887e-06, | |
| "loss": 0.0262, | |
| "step": 2425 | |
| }, | |
| { | |
| "epoch": 6.09, | |
| "learning_rate": 1.9864444444444446e-06, | |
| "loss": 0.0257, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 6.16, | |
| "learning_rate": 1.967e-06, | |
| "loss": 0.0287, | |
| "step": 2475 | |
| }, | |
| { | |
| "epoch": 6.22, | |
| "learning_rate": 1.9475555555555554e-06, | |
| "loss": 0.0272, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 6.28, | |
| "learning_rate": 1.9281111111111112e-06, | |
| "loss": 0.0269, | |
| "step": 2525 | |
| }, | |
| { | |
| "epoch": 6.34, | |
| "learning_rate": 1.9086666666666666e-06, | |
| "loss": 0.0256, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 6.41, | |
| "learning_rate": 1.8892222222222223e-06, | |
| "loss": 0.0279, | |
| "step": 2575 | |
| }, | |
| { | |
| "epoch": 6.47, | |
| "learning_rate": 1.8697777777777779e-06, | |
| "loss": 0.0253, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 6.53, | |
| "learning_rate": 1.850333333333333e-06, | |
| "loss": 0.0276, | |
| "step": 2625 | |
| }, | |
| { | |
| "epoch": 6.59, | |
| "learning_rate": 1.8308888888888887e-06, | |
| "loss": 0.0247, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 6.65, | |
| "learning_rate": 1.8114444444444443e-06, | |
| "loss": 0.0251, | |
| "step": 2675 | |
| }, | |
| { | |
| "epoch": 6.72, | |
| "learning_rate": 1.792e-06, | |
| "loss": 0.0248, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 6.78, | |
| "learning_rate": 1.7725555555555556e-06, | |
| "loss": 0.0248, | |
| "step": 2725 | |
| }, | |
| { | |
| "epoch": 6.84, | |
| "learning_rate": 1.7531111111111112e-06, | |
| "loss": 0.0222, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 6.9, | |
| "learning_rate": 1.7344444444444445e-06, | |
| "loss": 0.0253, | |
| "step": 2775 | |
| }, | |
| { | |
| "epoch": 6.97, | |
| "learning_rate": 1.7149999999999999e-06, | |
| "loss": 0.0223, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 7.03, | |
| "learning_rate": 1.6955555555555555e-06, | |
| "loss": 0.0208, | |
| "step": 2825 | |
| }, | |
| { | |
| "epoch": 7.09, | |
| "learning_rate": 1.676111111111111e-06, | |
| "loss": 0.0227, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 7.15, | |
| "learning_rate": 1.6566666666666665e-06, | |
| "loss": 0.021, | |
| "step": 2875 | |
| }, | |
| { | |
| "epoch": 7.21, | |
| "learning_rate": 1.6372222222222222e-06, | |
| "loss": 0.0213, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 7.28, | |
| "learning_rate": 1.6177777777777776e-06, | |
| "loss": 0.0212, | |
| "step": 2925 | |
| }, | |
| { | |
| "epoch": 7.34, | |
| "learning_rate": 1.5983333333333332e-06, | |
| "loss": 0.0199, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 7.4, | |
| "learning_rate": 1.5788888888888888e-06, | |
| "loss": 0.0213, | |
| "step": 2975 | |
| }, | |
| { | |
| "epoch": 7.46, | |
| "learning_rate": 1.5594444444444445e-06, | |
| "loss": 0.0195, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 7.46, | |
| "eval_loss": 0.2767290472984314, | |
| "eval_runtime": 1977.3759, | |
| "eval_samples_per_second": 0.858, | |
| "eval_steps_per_second": 0.107, | |
| "eval_wer": 9.881129271916791, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 7.52, | |
| "learning_rate": 1.54e-06, | |
| "loss": 0.0208, | |
| "step": 3025 | |
| }, | |
| { | |
| "epoch": 7.59, | |
| "learning_rate": 1.5205555555555557e-06, | |
| "loss": 0.0185, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 7.65, | |
| "learning_rate": 1.5011111111111111e-06, | |
| "loss": 0.0215, | |
| "step": 3075 | |
| }, | |
| { | |
| "epoch": 7.71, | |
| "learning_rate": 1.4816666666666667e-06, | |
| "loss": 0.0207, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 7.77, | |
| "learning_rate": 1.4622222222222224e-06, | |
| "loss": 0.0215, | |
| "step": 3125 | |
| }, | |
| { | |
| "epoch": 7.84, | |
| "learning_rate": 1.4427777777777778e-06, | |
| "loss": 0.0187, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 7.9, | |
| "learning_rate": 1.4233333333333334e-06, | |
| "loss": 0.0203, | |
| "step": 3175 | |
| }, | |
| { | |
| "epoch": 7.96, | |
| "learning_rate": 1.4038888888888888e-06, | |
| "loss": 0.02, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 8.02, | |
| "learning_rate": 1.3844444444444444e-06, | |
| "loss": 0.0178, | |
| "step": 3225 | |
| }, | |
| { | |
| "epoch": 8.08, | |
| "learning_rate": 1.365e-06, | |
| "loss": 0.0179, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 8.15, | |
| "learning_rate": 1.3455555555555555e-06, | |
| "loss": 0.0189, | |
| "step": 3275 | |
| }, | |
| { | |
| "epoch": 8.21, | |
| "learning_rate": 1.326111111111111e-06, | |
| "loss": 0.0168, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 8.27, | |
| "learning_rate": 1.3066666666666667e-06, | |
| "loss": 0.0188, | |
| "step": 3325 | |
| }, | |
| { | |
| "epoch": 8.33, | |
| "learning_rate": 1.2872222222222221e-06, | |
| "loss": 0.0167, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 8.4, | |
| "learning_rate": 1.2677777777777778e-06, | |
| "loss": 0.0159, | |
| "step": 3375 | |
| }, | |
| { | |
| "epoch": 8.46, | |
| "learning_rate": 1.2483333333333334e-06, | |
| "loss": 0.0175, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 8.52, | |
| "learning_rate": 1.2288888888888888e-06, | |
| "loss": 0.0156, | |
| "step": 3425 | |
| }, | |
| { | |
| "epoch": 8.58, | |
| "learning_rate": 1.2094444444444444e-06, | |
| "loss": 0.0177, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 8.64, | |
| "learning_rate": 1.19e-06, | |
| "loss": 0.0164, | |
| "step": 3475 | |
| }, | |
| { | |
| "epoch": 8.71, | |
| "learning_rate": 1.1705555555555555e-06, | |
| "loss": 0.0161, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 8.77, | |
| "learning_rate": 1.151111111111111e-06, | |
| "loss": 0.0147, | |
| "step": 3525 | |
| }, | |
| { | |
| "epoch": 8.83, | |
| "learning_rate": 1.1316666666666665e-06, | |
| "loss": 0.0184, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 8.89, | |
| "learning_rate": 1.1122222222222221e-06, | |
| "loss": 0.016, | |
| "step": 3575 | |
| }, | |
| { | |
| "epoch": 8.96, | |
| "learning_rate": 1.0927777777777777e-06, | |
| "loss": 0.0163, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 9.02, | |
| "learning_rate": 1.0733333333333332e-06, | |
| "loss": 0.018, | |
| "step": 3625 | |
| }, | |
| { | |
| "epoch": 9.08, | |
| "learning_rate": 1.0538888888888888e-06, | |
| "loss": 0.0158, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 9.14, | |
| "learning_rate": 1.0344444444444444e-06, | |
| "loss": 0.0148, | |
| "step": 3675 | |
| }, | |
| { | |
| "epoch": 9.2, | |
| "learning_rate": 1.0149999999999998e-06, | |
| "loss": 0.0143, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 9.27, | |
| "learning_rate": 9.955555555555554e-07, | |
| "loss": 0.0143, | |
| "step": 3725 | |
| }, | |
| { | |
| "epoch": 9.33, | |
| "learning_rate": 9.76111111111111e-07, | |
| "loss": 0.013, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 9.39, | |
| "learning_rate": 9.566666666666667e-07, | |
| "loss": 0.0149, | |
| "step": 3775 | |
| }, | |
| { | |
| "epoch": 9.45, | |
| "learning_rate": 9.372222222222222e-07, | |
| "loss": 0.0157, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 9.51, | |
| "learning_rate": 9.177777777777778e-07, | |
| "loss": 0.0155, | |
| "step": 3825 | |
| }, | |
| { | |
| "epoch": 9.58, | |
| "learning_rate": 8.983333333333332e-07, | |
| "loss": 0.0162, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 9.64, | |
| "learning_rate": 8.788888888888889e-07, | |
| "loss": 0.0138, | |
| "step": 3875 | |
| }, | |
| { | |
| "epoch": 9.7, | |
| "learning_rate": 8.594444444444444e-07, | |
| "loss": 0.0136, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 9.76, | |
| "learning_rate": 8.4e-07, | |
| "loss": 0.0144, | |
| "step": 3925 | |
| }, | |
| { | |
| "epoch": 9.83, | |
| "learning_rate": 8.205555555555556e-07, | |
| "loss": 0.015, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 9.89, | |
| "learning_rate": 8.011111111111111e-07, | |
| "loss": 0.0139, | |
| "step": 3975 | |
| }, | |
| { | |
| "epoch": 9.95, | |
| "learning_rate": 7.816666666666667e-07, | |
| "loss": 0.0126, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 9.95, | |
| "eval_loss": 0.28718188405036926, | |
| "eval_runtime": 1984.0507, | |
| "eval_samples_per_second": 0.855, | |
| "eval_steps_per_second": 0.107, | |
| "eval_wer": 9.50037147102526, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 10.01, | |
| "learning_rate": 7.622222222222222e-07, | |
| "loss": 0.0131, | |
| "step": 4025 | |
| }, | |
| { | |
| "epoch": 10.07, | |
| "learning_rate": 7.427777777777778e-07, | |
| "loss": 0.013, | |
| "step": 4050 | |
| }, | |
| { | |
| "epoch": 10.14, | |
| "learning_rate": 7.233333333333333e-07, | |
| "loss": 0.0139, | |
| "step": 4075 | |
| }, | |
| { | |
| "epoch": 10.2, | |
| "learning_rate": 7.038888888888888e-07, | |
| "loss": 0.0141, | |
| "step": 4100 | |
| }, | |
| { | |
| "epoch": 10.26, | |
| "learning_rate": 6.844444444444445e-07, | |
| "loss": 0.0116, | |
| "step": 4125 | |
| }, | |
| { | |
| "epoch": 10.32, | |
| "learning_rate": 6.65e-07, | |
| "loss": 0.0133, | |
| "step": 4150 | |
| }, | |
| { | |
| "epoch": 10.39, | |
| "learning_rate": 6.455555555555555e-07, | |
| "loss": 0.013, | |
| "step": 4175 | |
| }, | |
| { | |
| "epoch": 10.45, | |
| "learning_rate": 6.26111111111111e-07, | |
| "loss": 0.0114, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 10.51, | |
| "learning_rate": 6.066666666666666e-07, | |
| "loss": 0.0129, | |
| "step": 4225 | |
| }, | |
| { | |
| "epoch": 10.57, | |
| "learning_rate": 5.872222222222222e-07, | |
| "loss": 0.0117, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 10.63, | |
| "learning_rate": 5.677777777777777e-07, | |
| "loss": 0.0136, | |
| "step": 4275 | |
| }, | |
| { | |
| "epoch": 10.7, | |
| "learning_rate": 5.483333333333333e-07, | |
| "loss": 0.0126, | |
| "step": 4300 | |
| }, | |
| { | |
| "epoch": 10.76, | |
| "learning_rate": 5.288888888888889e-07, | |
| "loss": 0.0119, | |
| "step": 4325 | |
| }, | |
| { | |
| "epoch": 10.82, | |
| "learning_rate": 5.094444444444444e-07, | |
| "loss": 0.0141, | |
| "step": 4350 | |
| }, | |
| { | |
| "epoch": 10.88, | |
| "learning_rate": 4.900000000000001e-07, | |
| "loss": 0.0136, | |
| "step": 4375 | |
| }, | |
| { | |
| "epoch": 10.95, | |
| "learning_rate": 4.705555555555556e-07, | |
| "loss": 0.0123, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 11.01, | |
| "learning_rate": 4.511111111111111e-07, | |
| "loss": 0.012, | |
| "step": 4425 | |
| }, | |
| { | |
| "epoch": 11.07, | |
| "learning_rate": 4.316666666666667e-07, | |
| "loss": 0.0125, | |
| "step": 4450 | |
| }, | |
| { | |
| "epoch": 11.13, | |
| "learning_rate": 4.122222222222222e-07, | |
| "loss": 0.0109, | |
| "step": 4475 | |
| }, | |
| { | |
| "epoch": 11.19, | |
| "learning_rate": 3.9277777777777777e-07, | |
| "loss": 0.0117, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 11.26, | |
| "learning_rate": 3.7333333333333334e-07, | |
| "loss": 0.0128, | |
| "step": 4525 | |
| }, | |
| { | |
| "epoch": 11.32, | |
| "learning_rate": 3.5388888888888886e-07, | |
| "loss": 0.0112, | |
| "step": 4550 | |
| }, | |
| { | |
| "epoch": 11.38, | |
| "learning_rate": 3.3444444444444443e-07, | |
| "loss": 0.012, | |
| "step": 4575 | |
| }, | |
| { | |
| "epoch": 11.44, | |
| "learning_rate": 3.15e-07, | |
| "loss": 0.0115, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 11.5, | |
| "learning_rate": 2.9555555555555557e-07, | |
| "loss": 0.0116, | |
| "step": 4625 | |
| }, | |
| { | |
| "epoch": 11.57, | |
| "learning_rate": 2.761111111111111e-07, | |
| "loss": 0.0122, | |
| "step": 4650 | |
| }, | |
| { | |
| "epoch": 11.63, | |
| "learning_rate": 2.5666666666666666e-07, | |
| "loss": 0.0111, | |
| "step": 4675 | |
| }, | |
| { | |
| "epoch": 11.69, | |
| "learning_rate": 2.3722222222222223e-07, | |
| "loss": 0.0113, | |
| "step": 4700 | |
| }, | |
| { | |
| "epoch": 11.75, | |
| "learning_rate": 2.1777777777777778e-07, | |
| "loss": 0.0113, | |
| "step": 4725 | |
| }, | |
| { | |
| "epoch": 11.82, | |
| "learning_rate": 1.9833333333333332e-07, | |
| "loss": 0.0109, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 11.88, | |
| "learning_rate": 1.788888888888889e-07, | |
| "loss": 0.0128, | |
| "step": 4775 | |
| }, | |
| { | |
| "epoch": 11.94, | |
| "learning_rate": 1.5944444444444444e-07, | |
| "loss": 0.0102, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "learning_rate": 1.4e-07, | |
| "loss": 0.0115, | |
| "step": 4825 | |
| }, | |
| { | |
| "epoch": 12.06, | |
| "learning_rate": 1.2055555555555555e-07, | |
| "loss": 0.013, | |
| "step": 4850 | |
| }, | |
| { | |
| "epoch": 12.13, | |
| "learning_rate": 1.011111111111111e-07, | |
| "loss": 0.0104, | |
| "step": 4875 | |
| }, | |
| { | |
| "epoch": 12.19, | |
| "learning_rate": 8.166666666666667e-08, | |
| "loss": 0.0118, | |
| "step": 4900 | |
| }, | |
| { | |
| "epoch": 12.25, | |
| "learning_rate": 6.222222222222221e-08, | |
| "loss": 0.0108, | |
| "step": 4925 | |
| }, | |
| { | |
| "epoch": 12.31, | |
| "learning_rate": 4.277777777777778e-08, | |
| "loss": 0.0109, | |
| "step": 4950 | |
| }, | |
| { | |
| "epoch": 12.38, | |
| "learning_rate": 2.3333333333333334e-08, | |
| "loss": 0.0116, | |
| "step": 4975 | |
| }, | |
| { | |
| "epoch": 12.44, | |
| "learning_rate": 3.8888888888888884e-09, | |
| "loss": 0.0103, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 12.44, | |
| "eval_loss": 0.3021346628665924, | |
| "eval_runtime": 1982.8224, | |
| "eval_samples_per_second": 0.855, | |
| "eval_steps_per_second": 0.107, | |
| "eval_wer": 9.695393759286777, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 12.44, | |
| "step": 5000, | |
| "total_flos": 3.395544950145024e+20, | |
| "train_loss": 0.11761420190334321, | |
| "train_runtime": 43747.9022, | |
| "train_samples_per_second": 3.657, | |
| "train_steps_per_second": 0.114 | |
| } | |
| ], | |
| "max_steps": 5000, | |
| "num_train_epochs": 13, | |
| "total_flos": 3.395544950145024e+20, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |