wav2vec2-large-xlsr-mecita-coraa-portuguese-random-all-03
This model is a fine-tuned version of Edresson/wav2vec2-large-xlsr-coraa-portuguese on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.1364
- Wer: 0.0844
- Cer: 0.0267
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: 3e-05
- train_batch_size: 16
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 100
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
---|---|---|---|---|---|
29.1535 | 1.0 | 86 | 3.1676 | 1.0 | 1.0 |
7.924 | 2.0 | 172 | 2.9289 | 1.0 | 1.0 |
3.0115 | 3.0 | 258 | 2.8972 | 1.0 | 1.0 |
2.9304 | 4.0 | 344 | 2.8877 | 1.0 | 1.0 |
2.9073 | 5.0 | 430 | 2.8574 | 1.0 | 1.0 |
2.7919 | 6.0 | 516 | 1.9144 | 1.0 | 0.6742 |
1.6061 | 7.0 | 602 | 0.5439 | 0.2886 | 0.0766 |
1.6061 | 8.0 | 688 | 0.3726 | 0.1949 | 0.0558 |
0.7808 | 9.0 | 774 | 0.2960 | 0.1818 | 0.0517 |
0.5543 | 10.0 | 860 | 0.2591 | 0.1688 | 0.0477 |
0.4721 | 11.0 | 946 | 0.2367 | 0.1445 | 0.0427 |
0.414 | 12.0 | 1032 | 0.2167 | 0.1260 | 0.0376 |
0.3819 | 13.0 | 1118 | 0.1979 | 0.1150 | 0.0350 |
0.3376 | 14.0 | 1204 | 0.1877 | 0.1169 | 0.0346 |
0.3376 | 15.0 | 1290 | 0.1766 | 0.1084 | 0.0335 |
0.3199 | 16.0 | 1376 | 0.1754 | 0.1032 | 0.0323 |
0.3174 | 17.0 | 1462 | 0.1697 | 0.1017 | 0.0315 |
0.2747 | 18.0 | 1548 | 0.1668 | 0.0963 | 0.0308 |
0.2618 | 19.0 | 1634 | 0.1626 | 0.0937 | 0.0301 |
0.2557 | 20.0 | 1720 | 0.1597 | 0.0946 | 0.0299 |
0.2578 | 21.0 | 1806 | 0.1585 | 0.0944 | 0.0296 |
0.2578 | 22.0 | 1892 | 0.1549 | 0.0965 | 0.0302 |
0.2288 | 23.0 | 1978 | 0.1501 | 0.0939 | 0.0284 |
0.2269 | 24.0 | 2064 | 0.1524 | 0.0944 | 0.0291 |
0.2327 | 25.0 | 2150 | 0.1476 | 0.0903 | 0.0281 |
0.2024 | 26.0 | 2236 | 0.1481 | 0.0903 | 0.0284 |
0.2056 | 27.0 | 2322 | 0.1434 | 0.0925 | 0.0284 |
0.2097 | 28.0 | 2408 | 0.1468 | 0.0894 | 0.0280 |
0.2097 | 29.0 | 2494 | 0.1435 | 0.0860 | 0.0273 |
0.2177 | 30.0 | 2580 | 0.1498 | 0.0877 | 0.0281 |
0.1935 | 31.0 | 2666 | 0.1452 | 0.0891 | 0.0278 |
0.1918 | 32.0 | 2752 | 0.1466 | 0.0849 | 0.0275 |
0.1805 | 33.0 | 2838 | 0.1437 | 0.0889 | 0.0282 |
0.1805 | 34.0 | 2924 | 0.1409 | 0.0870 | 0.0274 |
0.1835 | 35.0 | 3010 | 0.1422 | 0.0856 | 0.0271 |
0.1835 | 36.0 | 3096 | 0.1377 | 0.0851 | 0.0264 |
0.1787 | 37.0 | 3182 | 0.1364 | 0.0844 | 0.0267 |
0.1695 | 38.0 | 3268 | 0.1418 | 0.0849 | 0.0268 |
0.1775 | 39.0 | 3354 | 0.1401 | 0.0844 | 0.0270 |
0.1763 | 40.0 | 3440 | 0.1402 | 0.0815 | 0.0265 |
0.1702 | 41.0 | 3526 | 0.1418 | 0.0830 | 0.0264 |
0.1569 | 42.0 | 3612 | 0.1400 | 0.0825 | 0.0258 |
0.1569 | 43.0 | 3698 | 0.1401 | 0.0815 | 0.0262 |
0.1617 | 44.0 | 3784 | 0.1406 | 0.0792 | 0.0262 |
0.1596 | 45.0 | 3870 | 0.1395 | 0.0818 | 0.0264 |
0.1431 | 46.0 | 3956 | 0.1382 | 0.0815 | 0.0262 |
0.158 | 47.0 | 4042 | 0.1391 | 0.0813 | 0.0265 |
0.1552 | 48.0 | 4128 | 0.1393 | 0.0825 | 0.0266 |
0.1379 | 49.0 | 4214 | 0.1371 | 0.0811 | 0.0256 |
0.145 | 50.0 | 4300 | 0.1392 | 0.0801 | 0.0256 |
0.145 | 51.0 | 4386 | 0.1416 | 0.0820 | 0.0262 |
0.1647 | 52.0 | 4472 | 0.1392 | 0.0789 | 0.0256 |
0.1493 | 53.0 | 4558 | 0.1425 | 0.0794 | 0.0257 |
0.1492 | 54.0 | 4644 | 0.1419 | 0.0796 | 0.0257 |
0.139 | 55.0 | 4730 | 0.1400 | 0.0758 | 0.0250 |
0.1385 | 56.0 | 4816 | 0.1424 | 0.0792 | 0.0253 |
0.128 | 57.0 | 4902 | 0.1403 | 0.0806 | 0.0253 |
Framework versions
- Transformers 4.28.0
- Pytorch 2.5.0+cu121
- Datasets 3.1.0
- Tokenizers 0.13.3
- Downloads last month
- 79