w2v2_ablation_with_ling_head-0drop-load-best-per-best_on_tp0.025_tl10_fp0.001_fl16
This model is a fine-tuned version of nguyenvulebinh/wav2vec2-base-vietnamese-250h on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.4603
- Wer: 0.1849
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
More information needed
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 2e-05
- train_batch_size: 8
- eval_batch_size: 32
- seed: 42
- distributed_type: multi-GPU
- num_devices: 4
- total_train_batch_size: 32
- total_eval_batch_size: 128
- optimizer: Adam with betas=(0.9,0.98) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 100
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Wer |
---|---|---|---|---|
114.032 | 0.94 | 100 | 98.5210 | 21.4967 |
63.4912 | 1.89 | 200 | 9.8990 | 1.0 |
5.9426 | 2.83 | 300 | 5.2909 | 1.0 |
5.0183 | 3.77 | 400 | 5.2482 | 1.0 |
4.6782 | 4.72 | 500 | 5.5314 | 1.0 |
4.4732 | 5.66 | 600 | 5.2250 | 1.0 |
4.4059 | 6.6 | 700 | 5.1483 | 1.0 |
4.3368 | 7.55 | 800 | 4.9275 | 1.0 |
4.2178 | 8.49 | 900 | 4.8987 | 1.0 |
3.913 | 9.43 | 1000 | 3.7007 | 0.8807 |
2.7998 | 10.38 | 1100 | 2.1296 | 0.5331 |
1.8405 | 11.32 | 1200 | 1.4873 | 0.4636 |
1.2987 | 12.26 | 1300 | 1.0532 | 0.3338 |
1.0387 | 13.21 | 1400 | 0.8759 | 0.3348 |
0.851 | 14.15 | 1500 | 0.7743 | 0.3604 |
0.7128 | 15.09 | 1600 | 0.6523 | 0.2796 |
0.605 | 16.04 | 1700 | 0.6352 | 0.2995 |
0.5315 | 16.98 | 1800 | 0.5920 | 0.2603 |
0.4845 | 17.92 | 1900 | 0.5476 | 0.2503 |
0.4257 | 18.87 | 2000 | 0.5398 | 0.2285 |
0.4124 | 19.81 | 2100 | 0.5378 | 0.2764 |
0.3595 | 20.75 | 2200 | 0.5109 | 0.2147 |
0.3958 | 21.7 | 2300 | 0.4825 | 0.2342 |
0.3546 | 22.64 | 2400 | 0.4649 | 0.2251 |
0.304 | 23.58 | 2500 | 0.4701 | 0.2115 |
0.291 | 24.53 | 2600 | 0.4515 | 0.2180 |
0.2946 | 25.47 | 2700 | 0.4537 | 0.2012 |
0.2588 | 26.42 | 2800 | 0.4423 | 0.1939 |
0.2625 | 27.36 | 2900 | 0.4493 | 0.1924 |
0.2385 | 28.3 | 3000 | 0.4364 | 0.1724 |
0.2327 | 29.25 | 3100 | 0.4382 | 0.1967 |
0.26 | 30.19 | 3200 | 0.4454 | 0.1823 |
0.2151 | 31.13 | 3300 | 0.4424 | 0.1987 |
0.2213 | 32.08 | 3400 | 0.4377 | 0.2085 |
0.2226 | 33.02 | 3500 | 0.4375 | 0.2095 |
0.208 | 33.96 | 3600 | 0.4358 | 0.1994 |
0.2061 | 34.91 | 3700 | 0.4308 | 0.1919 |
0.1929 | 35.85 | 3800 | 0.4298 | 0.1905 |
0.1786 | 36.79 | 3900 | 0.4139 | 0.1974 |
0.172 | 37.74 | 4000 | 0.4183 | 0.1823 |
0.1769 | 38.68 | 4100 | 0.4252 | 0.1890 |
0.1813 | 39.62 | 4200 | 0.4360 | 0.1880 |
0.1676 | 40.57 | 4300 | 0.4325 | 0.1770 |
0.1581 | 41.51 | 4400 | 0.4386 | 0.1755 |
0.17 | 42.45 | 4500 | 0.4374 | 0.1979 |
0.1778 | 43.4 | 4600 | 0.4360 | 0.1726 |
0.162 | 44.34 | 4700 | 0.4424 | 0.1822 |
0.1605 | 45.28 | 4800 | 0.4500 | 0.2065 |
0.1472 | 46.23 | 4900 | 0.4555 | 0.2102 |
0.1428 | 47.17 | 5000 | 0.4358 | 0.1733 |
0.1393 | 48.11 | 5100 | 0.4406 | 0.1904 |
0.1444 | 49.06 | 5200 | 0.4481 | 0.2030 |
0.1401 | 50.0 | 5300 | 0.4507 | 0.1952 |
0.1311 | 50.94 | 5400 | 0.4353 | 0.1857 |
0.1337 | 51.89 | 5500 | 0.4439 | 0.2018 |
0.1289 | 52.83 | 5600 | 0.4461 | 0.1946 |
0.1234 | 53.77 | 5700 | 0.4395 | 0.2048 |
0.1301 | 54.72 | 5800 | 0.4590 | 0.2114 |
0.1378 | 55.66 | 5900 | 0.4548 | 0.2144 |
0.1251 | 56.6 | 6000 | 0.4477 | 0.1877 |
0.1224 | 57.55 | 6100 | 0.4478 | 0.1933 |
0.1233 | 58.49 | 6200 | 0.4467 | 0.1841 |
0.1237 | 59.43 | 6300 | 0.4399 | 0.1834 |
0.1176 | 60.38 | 6400 | 0.4471 | 0.2097 |
0.1117 | 61.32 | 6500 | 0.4587 | 0.1970 |
0.111 | 62.26 | 6600 | 0.4707 | 0.2102 |
0.1239 | 63.21 | 6700 | 0.4518 | 0.1923 |
0.1152 | 64.15 | 6800 | 0.4503 | 0.1967 |
0.1121 | 65.09 | 6900 | 0.4467 | 0.1944 |
0.1175 | 66.04 | 7000 | 0.4486 | 0.1914 |
0.1242 | 66.98 | 7100 | 0.4537 | 0.1973 |
0.111 | 67.92 | 7200 | 0.4587 | 0.2008 |
0.1063 | 68.87 | 7300 | 0.4551 | 0.1929 |
0.1133 | 69.81 | 7400 | 0.4547 | 0.1929 |
0.1098 | 70.75 | 7500 | 0.4512 | 0.1982 |
0.1123 | 71.7 | 7600 | 0.4578 | 0.1955 |
0.1144 | 72.64 | 7700 | 0.4533 | 0.1830 |
0.1113 | 73.58 | 7800 | 0.4545 | 0.1788 |
0.0968 | 74.53 | 7900 | 0.4584 | 0.1725 |
0.0951 | 75.47 | 8000 | 0.4646 | 0.1859 |
0.0982 | 76.42 | 8100 | 0.4557 | 0.1813 |
0.0959 | 77.36 | 8200 | 0.4566 | 0.1742 |
0.093 | 78.3 | 8300 | 0.4604 | 0.1880 |
0.103 | 79.25 | 8400 | 0.4614 | 0.1908 |
0.1101 | 80.19 | 8500 | 0.4586 | 0.1805 |
0.1046 | 81.13 | 8600 | 0.4590 | 0.1825 |
0.0979 | 82.08 | 8700 | 0.4555 | 0.1762 |
0.103 | 83.02 | 8800 | 0.4573 | 0.1780 |
0.0958 | 83.96 | 8900 | 0.4575 | 0.1803 |
0.0948 | 84.91 | 9000 | 0.4581 | 0.1814 |
0.1003 | 85.85 | 9100 | 0.4600 | 0.1830 |
0.1066 | 86.79 | 9200 | 0.4609 | 0.1870 |
0.0887 | 87.74 | 9300 | 0.4615 | 0.1834 |
0.0936 | 88.68 | 9400 | 0.4610 | 0.1819 |
0.0892 | 89.62 | 9500 | 0.4595 | 0.1801 |
0.1039 | 90.57 | 9600 | 0.4612 | 0.1837 |
0.097 | 91.51 | 9700 | 0.4610 | 0.1834 |
0.0969 | 92.45 | 9800 | 0.4605 | 0.1844 |
0.0946 | 93.4 | 9900 | 0.4596 | 0.1843 |
0.0947 | 94.34 | 10000 | 0.4605 | 0.1850 |
0.095 | 95.28 | 10100 | 0.4616 | 0.1861 |
0.0856 | 96.23 | 10200 | 0.4611 | 0.1853 |
0.0983 | 97.17 | 10300 | 0.4603 | 0.1850 |
0.0947 | 98.11 | 10400 | 0.4605 | 0.1853 |
0.0948 | 99.06 | 10500 | 0.4604 | 0.1853 |
0.0917 | 100.0 | 10600 | 0.4603 | 0.1849 |
Framework versions
- Transformers 4.35.2
- Pytorch 1.13.1+cu117
- Datasets 2.12.0
- Tokenizers 0.14.1
- Downloads last month
- 8
Model tree for tuanio/w2v2_ablation_with_ling_head-0drop-load-best-per-best_on_tp0.025_tl10_fp0.001_fl16
Base model
nguyenvulebinh/wav2vec2-base-vietnamese-250h