<!-- 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-grade-2-3-5
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: 0.1410
- Wer: 0.0830
- Cer: 0.0249
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 |
---|---|---|---|---|---|
23.7508 | 1.0 | 79 | 3.2381 | 1.0 | 1.0 |
7.0931 | 2.0 | 158 | 2.9568 | 1.0 | 1.0 |
3.0421 | 3.0 | 237 | 2.9135 | 1.0 | 1.0 |
2.9245 | 4.0 | 316 | 2.9112 | 1.0 | 1.0 |
2.9245 | 5.0 | 395 | 2.9111 | 1.0 | 1.0 |
2.9084 | 6.0 | 474 | 2.1442 | 1.0 | 0.7405 |
2.5789 | 7.0 | 553 | 0.6584 | 0.3189 | 0.0840 |
1.1666 | 8.0 | 632 | 0.4341 | 0.2224 | 0.0595 |
0.7078 | 9.0 | 711 | 0.3466 | 0.1869 | 0.0520 |
0.7078 | 10.0 | 790 | 0.2976 | 0.1601 | 0.0461 |
0.5229 | 11.0 | 869 | 0.2709 | 0.1461 | 0.0426 |
0.4841 | 12.0 | 948 | 0.2418 | 0.1295 | 0.0383 |
0.4215 | 13.0 | 1027 | 0.2325 | 0.1265 | 0.0375 |
0.3987 | 14.0 | 1106 | 0.2156 | 0.1130 | 0.0339 |
0.3987 | 15.0 | 1185 | 0.2075 | 0.1103 | 0.0331 |
0.348 | 16.0 | 1264 | 0.2015 | 0.1072 | 0.0331 |
0.3193 | 17.0 | 1343 | 0.1882 | 0.1028 | 0.0309 |
0.2812 | 18.0 | 1422 | 0.1860 | 0.1020 | 0.0307 |
0.2726 | 19.0 | 1501 | 0.1803 | 0.0990 | 0.0302 |
0.2726 | 20.0 | 1580 | 0.1775 | 0.1023 | 0.0298 |
0.2715 | 21.0 | 1659 | 0.1725 | 0.1006 | 0.0297 |
0.2767 | 22.0 | 1738 | 0.1724 | 0.0970 | 0.0290 |
0.232 | 23.0 | 1817 | 0.1632 | 0.0992 | 0.0288 |
0.232 | 24.0 | 1896 | 0.1647 | 0.0940 | 0.0284 |
0.2357 | 25.0 | 1975 | 0.1628 | 0.0945 | 0.0279 |
0.2191 | 26.0 | 2054 | 0.1627 | 0.0923 | 0.0276 |
0.2176 | 27.0 | 2133 | 0.1591 | 0.0918 | 0.0278 |
0.2046 | 28.0 | 2212 | 0.1552 | 0.0915 | 0.0268 |
0.2046 | 29.0 | 2291 | 0.1483 | 0.0907 | 0.0269 |
0.2196 | 30.0 | 2370 | 0.1458 | 0.0890 | 0.0263 |
0.1999 | 31.0 | 2449 | 0.1500 | 0.0896 | 0.0268 |
0.1877 | 32.0 | 2528 | 0.1505 | 0.0885 | 0.0262 |
0.181 | 33.0 | 2607 | 0.1537 | 0.0885 | 0.0265 |
0.181 | 34.0 | 2686 | 0.1532 | 0.0843 | 0.0254 |
0.1754 | 35.0 | 2765 | 0.1458 | 0.0843 | 0.0255 |
0.1734 | 36.0 | 2844 | 0.1458 | 0.0874 | 0.0258 |
0.1843 | 37.0 | 2923 | 0.1422 | 0.0835 | 0.0252 |
0.1898 | 38.0 | 3002 | 0.1466 | 0.0863 | 0.0255 |
0.1898 | 39.0 | 3081 | 0.1495 | 0.0821 | 0.0250 |
0.162 | 40.0 | 3160 | 0.1489 | 0.0830 | 0.0253 |
0.1709 | 41.0 | 3239 | 0.1432 | 0.0799 | 0.0246 |
0.1584 | 42.0 | 3318 | 0.1432 | 0.0830 | 0.0248 |
0.1584 | 43.0 | 3397 | 0.1410 | 0.0830 | 0.0249 |
0.1626 | 44.0 | 3476 | 0.1463 | 0.0890 | 0.0261 |
0.1757 | 45.0 | 3555 | 0.1442 | 0.0819 | 0.0250 |
0.169 | 46.0 | 3634 | 0.1492 | 0.0860 | 0.0258 |
0.1542 | 47.0 | 3713 | 0.1488 | 0.0843 | 0.0252 |
0.1542 | 48.0 | 3792 | 0.1478 | 0.0832 | 0.0248 |
0.1476 | 49.0 | 3871 | 0.1461 | 0.0857 | 0.0250 |
0.1718 | 50.0 | 3950 | 0.1466 | 0.0838 | 0.0245 |
0.1395 | 51.0 | 4029 | 0.1449 | 0.0849 | 0.0248 |
0.1433 | 52.0 | 4108 | 0.1430 | 0.0813 | 0.0243 |
0.1433 | 53.0 | 4187 | 0.1461 | 0.0827 | 0.0250 |
0.1525 | 54.0 | 4266 | 0.1455 | 0.0838 | 0.0244 |
0.141 | 55.0 | 4345 | 0.1502 | 0.0857 | 0.0248 |
0.1395 | 56.0 | 4424 | 0.1472 | 0.0843 | 0.0249 |
0.1405 | 57.0 | 4503 | 0.1457 | 0.0843 | 0.0245 |
0.1405 | 58.0 | 4582 | 0.1487 | 0.0835 | 0.0247 |
0.1385 | 59.0 | 4661 | 0.1452 | 0.0835 | 0.0246 |
0.1527 | 60.0 | 4740 | 0.1502 | 0.0827 | 0.0246 |
0.133 | 61.0 | 4819 | 0.1460 | 0.0827 | 0.0246 |
0.133 | 62.0 | 4898 | 0.1487 | 0.0824 | 0.0247 |
0.1394 | 63.0 | 4977 | 0.1471 | 0.0821 | 0.0246 |
Framework versions
- Transformers 4.28.0
- Pytorch 2.0.1+cu117
- Datasets 2.14.4
- Tokenizers 0.13.3