<!-- 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-mec-ita-coraa-portuguese-all-07
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: 0.2122
- Wer: 0.0947
- Cer: 0.0328
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 |
|---|---|---|---|---|---|
| 30.5249 | 1.0 | 86 | 3.3024 | 1.0 | 1.0 |
| 7.7867 | 2.0 | 172 | 2.9628 | 1.0 | 1.0 |
| 3.0424 | 3.0 | 258 | 2.9183 | 1.0 | 1.0 |
| 2.9279 | 4.0 | 344 | 2.8627 | 1.0 | 1.0 |
| 2.782 | 5.0 | 430 | 1.2256 | 0.9193 | 0.2503 |
| 1.4877 | 6.0 | 516 | 0.5729 | 0.2641 | 0.0747 |
| 0.8015 | 7.0 | 602 | 0.4157 | 0.1947 | 0.0561 |
| 0.8015 | 8.0 | 688 | 0.3609 | 0.1720 | 0.0513 |
| 0.6062 | 9.0 | 774 | 0.3232 | 0.1593 | 0.0477 |
| 0.5305 | 10.0 | 860 | 0.3114 | 0.1516 | 0.0463 |
| 0.4351 | 11.0 | 946 | 0.2865 | 0.1296 | 0.0411 |
| 0.4126 | 12.0 | 1032 | 0.2778 | 0.1267 | 0.0404 |
| 0.3665 | 13.0 | 1118 | 0.2677 | 0.1207 | 0.0391 |
| 0.3415 | 14.0 | 1204 | 0.2625 | 0.1231 | 0.0395 |
| 0.3415 | 15.0 | 1290 | 0.2548 | 0.1154 | 0.0382 |
| 0.3013 | 16.0 | 1376 | 0.2517 | 0.1089 | 0.0367 |
| 0.2869 | 17.0 | 1462 | 0.2377 | 0.1041 | 0.0362 |
| 0.2908 | 18.0 | 1548 | 0.2423 | 0.1014 | 0.0351 |
| 0.2815 | 19.0 | 1634 | 0.2452 | 0.1043 | 0.0357 |
| 0.262 | 20.0 | 1720 | 0.2397 | 0.1024 | 0.0360 |
| 0.2456 | 21.0 | 1806 | 0.2197 | 0.0986 | 0.0342 |
| 0.2456 | 22.0 | 1892 | 0.2307 | 0.0973 | 0.0340 |
| 0.2421 | 23.0 | 1978 | 0.2255 | 0.0976 | 0.0336 |
| 0.2298 | 24.0 | 2064 | 0.2188 | 0.0966 | 0.0332 |
| 0.221 | 25.0 | 2150 | 0.2167 | 0.0935 | 0.0327 |
| 0.2123 | 26.0 | 2236 | 0.2228 | 0.0908 | 0.0312 |
| 0.2182 | 27.0 | 2322 | 0.2154 | 0.0913 | 0.0322 |
| 0.2041 | 28.0 | 2408 | 0.2326 | 0.0928 | 0.0330 |
| 0.2041 | 29.0 | 2494 | 0.2191 | 0.0920 | 0.0320 |
| 0.1971 | 30.0 | 2580 | 0.2241 | 0.0940 | 0.0330 |
| 0.2073 | 31.0 | 2666 | 0.2122 | 0.0947 | 0.0328 |
| 0.2086 | 32.0 | 2752 | 0.2217 | 0.0918 | 0.0313 |
| 0.1828 | 33.0 | 2838 | 0.2163 | 0.0920 | 0.0312 |
| 0.1939 | 34.0 | 2924 | 0.2186 | 0.0952 | 0.0323 |
| 0.1944 | 35.0 | 3010 | 0.2199 | 0.0884 | 0.0306 |
| 0.1944 | 36.0 | 3096 | 0.2127 | 0.0913 | 0.0309 |
| 0.1793 | 37.0 | 3182 | 0.2206 | 0.0899 | 0.0299 |
| 0.1843 | 38.0 | 3268 | 0.2190 | 0.0901 | 0.0305 |
| 0.1875 | 39.0 | 3354 | 0.2215 | 0.0899 | 0.0306 |
| 0.1587 | 40.0 | 3440 | 0.2215 | 0.0911 | 0.0307 |
| 0.1575 | 41.0 | 3526 | 0.2179 | 0.0872 | 0.0304 |
| 0.1853 | 42.0 | 3612 | 0.2183 | 0.0877 | 0.0302 |
| 0.1853 | 43.0 | 3698 | 0.2165 | 0.0865 | 0.0300 |
| 0.1696 | 44.0 | 3784 | 0.2196 | 0.0896 | 0.0305 |
| 0.1507 | 45.0 | 3870 | 0.2160 | 0.0884 | 0.0304 |
| 0.1509 | 46.0 | 3956 | 0.2163 | 0.0882 | 0.0299 |
| 0.1507 | 47.0 | 4042 | 0.2175 | 0.0896 | 0.0307 |
| 0.1596 | 48.0 | 4128 | 0.2234 | 0.0894 | 0.0305 |
| 0.1487 | 49.0 | 4214 | 0.2294 | 0.0877 | 0.0302 |
| 0.1512 | 50.0 | 4300 | 0.2180 | 0.0882 | 0.0306 |
| 0.1512 | 51.0 | 4386 | 0.2354 | 0.0858 | 0.0299 |
Framework versions
- Transformers 4.28.1
- Pytorch 2.0.0+cu118
- Datasets 2.11.0
- Tokenizers 0.13.3
