<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->
wav2vec2-large-xlsr-mecita-coraa-portuguese-all-clean-06
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: 0.1240
- Wer: 0.0822
- Cer: 0.0258
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 |
---|---|---|---|---|---|
25.6341 | 1.0 | 67 | 3.3417 | 1.0 | 1.0 |
7.4924 | 2.0 | 134 | 2.9735 | 1.0 | 1.0 |
3.0117 | 3.0 | 201 | 2.9203 | 1.0 | 1.0 |
3.0117 | 4.0 | 268 | 2.9038 | 1.0 | 1.0 |
2.9154 | 5.0 | 335 | 2.8627 | 1.0 | 1.0 |
2.7879 | 6.0 | 402 | 2.1177 | 1.0 | 0.6809 |
2.7879 | 7.0 | 469 | 0.6995 | 0.3645 | 0.0918 |
1.4902 | 8.0 | 536 | 0.4364 | 0.2452 | 0.0628 |
0.7612 | 9.0 | 603 | 0.3263 | 0.1858 | 0.0500 |
0.7612 | 10.0 | 670 | 0.2701 | 0.1576 | 0.0437 |
0.576 | 11.0 | 737 | 0.2434 | 0.1512 | 0.0411 |
0.4776 | 12.0 | 804 | 0.2220 | 0.1389 | 0.0392 |
0.4776 | 13.0 | 871 | 0.2046 | 0.1318 | 0.0368 |
0.4025 | 14.0 | 938 | 0.1913 | 0.1158 | 0.0338 |
0.3331 | 15.0 | 1005 | 0.1784 | 0.1124 | 0.0328 |
0.3331 | 16.0 | 1072 | 0.1682 | 0.1077 | 0.0320 |
0.3441 | 17.0 | 1139 | 0.1603 | 0.1033 | 0.0301 |
0.3075 | 18.0 | 1206 | 0.1527 | 0.0975 | 0.0295 |
0.3075 | 19.0 | 1273 | 0.1499 | 0.1077 | 0.0299 |
0.2895 | 20.0 | 1340 | 0.1416 | 0.0992 | 0.0289 |
0.2512 | 21.0 | 1407 | 0.1400 | 0.0985 | 0.0292 |
0.2512 | 22.0 | 1474 | 0.1389 | 0.0934 | 0.0290 |
0.2711 | 23.0 | 1541 | 0.1339 | 0.0910 | 0.0272 |
0.2491 | 24.0 | 1608 | 0.1357 | 0.0917 | 0.0280 |
0.2491 | 25.0 | 1675 | 0.1328 | 0.0924 | 0.0282 |
0.2176 | 26.0 | 1742 | 0.1320 | 0.0883 | 0.0267 |
0.2353 | 27.0 | 1809 | 0.1338 | 0.0904 | 0.0280 |
0.2353 | 28.0 | 1876 | 0.1310 | 0.0849 | 0.0273 |
0.2177 | 29.0 | 1943 | 0.1240 | 0.0897 | 0.0266 |
0.1974 | 30.0 | 2010 | 0.1284 | 0.0859 | 0.0262 |
0.1974 | 31.0 | 2077 | 0.1309 | 0.0856 | 0.0262 |
0.1956 | 32.0 | 2144 | 0.1296 | 0.0893 | 0.0269 |
0.2159 | 33.0 | 2211 | 0.1305 | 0.0842 | 0.0259 |
0.2159 | 34.0 | 2278 | 0.1281 | 0.0849 | 0.0258 |
0.1882 | 35.0 | 2345 | 0.1284 | 0.0853 | 0.0263 |
0.1833 | 36.0 | 2412 | 0.1289 | 0.0863 | 0.0261 |
0.1833 | 37.0 | 2479 | 0.1283 | 0.0907 | 0.0274 |
0.1646 | 38.0 | 2546 | 0.1282 | 0.0921 | 0.0274 |
0.1745 | 39.0 | 2613 | 0.1263 | 0.0815 | 0.0260 |
0.1745 | 40.0 | 2680 | 0.1308 | 0.0842 | 0.0263 |
0.1891 | 41.0 | 2747 | 0.1287 | 0.0812 | 0.0262 |
0.1685 | 42.0 | 2814 | 0.1300 | 0.0825 | 0.0259 |
0.1685 | 43.0 | 2881 | 0.1240 | 0.0822 | 0.0258 |
0.1569 | 44.0 | 2948 | 0.1301 | 0.0839 | 0.0263 |
0.1733 | 45.0 | 3015 | 0.1277 | 0.0808 | 0.0256 |
0.1733 | 46.0 | 3082 | 0.1268 | 0.0808 | 0.0254 |
0.162 | 47.0 | 3149 | 0.1263 | 0.0802 | 0.0250 |
0.1446 | 48.0 | 3216 | 0.1245 | 0.0795 | 0.0247 |
0.1446 | 49.0 | 3283 | 0.1249 | 0.0785 | 0.0249 |
0.1459 | 50.0 | 3350 | 0.1254 | 0.0825 | 0.0255 |
0.1497 | 51.0 | 3417 | 0.1272 | 0.0815 | 0.0256 |
0.1497 | 52.0 | 3484 | 0.1253 | 0.0822 | 0.0250 |
0.1359 | 53.0 | 3551 | 0.1312 | 0.0805 | 0.0251 |
0.1544 | 54.0 | 3618 | 0.1333 | 0.0802 | 0.0253 |
0.1544 | 55.0 | 3685 | 0.1319 | 0.0832 | 0.0260 |
0.1382 | 56.0 | 3752 | 0.1330 | 0.0832 | 0.0258 |
0.1451 | 57.0 | 3819 | 0.1313 | 0.0815 | 0.0251 |
0.1451 | 58.0 | 3886 | 0.1310 | 0.0815 | 0.0254 |
0.1341 | 59.0 | 3953 | 0.1341 | 0.0785 | 0.0251 |
0.1498 | 60.0 | 4020 | 0.1296 | 0.0849 | 0.0260 |
0.1498 | 61.0 | 4087 | 0.1303 | 0.0785 | 0.0250 |
0.1435 | 62.0 | 4154 | 0.1280 | 0.0798 | 0.0252 |
0.1407 | 63.0 | 4221 | 0.1301 | 0.0815 | 0.0248 |
Framework versions
- Transformers 4.28.0
- Pytorch 2.0.0+cu117
- Datasets 2.12.0
- Tokenizers 0.13.3