<!-- 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-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: 0.0848
- Wer: 0.0736
- Cer: 0.0202
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 |
---|---|---|---|---|---|
18.9736 | 0.99 | 38 | 3.8017 | 1.0 | 1.0 |
18.9736 | 2.0 | 77 | 3.1260 | 1.0 | 1.0 |
6.1371 | 2.99 | 115 | 2.9654 | 1.0 | 1.0 |
6.1371 | 4.0 | 154 | 2.9220 | 1.0 | 1.0 |
6.1371 | 4.99 | 192 | 2.8953 | 1.0 | 1.0 |
2.9716 | 6.0 | 231 | 2.8901 | 1.0 | 1.0 |
2.9716 | 6.99 | 269 | 2.8890 | 1.0 | 1.0 |
2.9025 | 8.0 | 308 | 2.8958 | 1.0 | 1.0 |
2.9025 | 8.99 | 346 | 2.8737 | 1.0 | 1.0 |
2.9025 | 10.0 | 385 | 2.8267 | 1.0 | 1.0 |
2.8651 | 10.99 | 423 | 2.5487 | 0.9973 | 0.9974 |
2.8651 | 12.0 | 462 | 1.5574 | 0.9960 | 0.4254 |
2.1616 | 12.99 | 500 | 0.7383 | 0.3434 | 0.0847 |
2.1616 | 14.0 | 539 | 0.4702 | 0.2181 | 0.0557 |
2.1616 | 14.99 | 577 | 0.3628 | 0.1858 | 0.0468 |
0.8473 | 16.0 | 616 | 0.2965 | 0.1661 | 0.0416 |
0.8473 | 16.99 | 654 | 0.2573 | 0.1504 | 0.0386 |
0.8473 | 18.0 | 693 | 0.2312 | 0.1436 | 0.0367 |
0.5711 | 18.99 | 731 | 0.2087 | 0.1378 | 0.0347 |
0.5711 | 20.0 | 770 | 0.1902 | 0.1315 | 0.0340 |
0.4431 | 20.99 | 808 | 0.1767 | 0.1252 | 0.0326 |
0.4431 | 22.0 | 847 | 0.1643 | 0.1154 | 0.0299 |
0.4431 | 22.99 | 885 | 0.1589 | 0.1037 | 0.0284 |
0.3674 | 24.0 | 924 | 0.1505 | 0.0956 | 0.0266 |
0.3674 | 24.99 | 962 | 0.1447 | 0.1005 | 0.0269 |
0.3316 | 26.0 | 1001 | 0.1370 | 0.0911 | 0.0252 |
0.3316 | 26.99 | 1039 | 0.1345 | 0.0889 | 0.0247 |
0.3316 | 28.0 | 1078 | 0.1316 | 0.0853 | 0.0244 |
0.2824 | 28.99 | 1116 | 0.1253 | 0.0839 | 0.0238 |
0.2824 | 30.0 | 1155 | 0.1213 | 0.0794 | 0.0230 |
0.2824 | 30.99 | 1193 | 0.1183 | 0.0830 | 0.0235 |
0.261 | 32.0 | 1232 | 0.1172 | 0.0790 | 0.0228 |
0.261 | 32.99 | 1270 | 0.1148 | 0.0808 | 0.0232 |
0.241 | 34.0 | 1309 | 0.1136 | 0.0817 | 0.0241 |
0.241 | 34.99 | 1347 | 0.1098 | 0.0768 | 0.0226 |
0.241 | 36.0 | 1386 | 0.1081 | 0.0768 | 0.0218 |
0.2247 | 36.99 | 1424 | 0.1076 | 0.0781 | 0.0227 |
0.2247 | 38.0 | 1463 | 0.1063 | 0.0781 | 0.0224 |
0.2219 | 38.99 | 1501 | 0.1045 | 0.0750 | 0.0213 |
0.2219 | 40.0 | 1540 | 0.1064 | 0.0745 | 0.0223 |
0.2219 | 40.99 | 1578 | 0.1037 | 0.0750 | 0.0222 |
0.2172 | 42.0 | 1617 | 0.1027 | 0.0736 | 0.0220 |
0.2172 | 42.99 | 1655 | 0.1008 | 0.0741 | 0.0219 |
0.2172 | 44.0 | 1694 | 0.0995 | 0.0741 | 0.0214 |
0.2169 | 44.99 | 1732 | 0.0986 | 0.0714 | 0.0211 |
0.2169 | 46.0 | 1771 | 0.0967 | 0.0745 | 0.0211 |
0.2044 | 46.99 | 1809 | 0.0953 | 0.0763 | 0.0220 |
0.2044 | 48.0 | 1848 | 0.0944 | 0.0727 | 0.0214 |
0.2044 | 48.99 | 1886 | 0.0936 | 0.0736 | 0.0212 |
0.1929 | 50.0 | 1925 | 0.0923 | 0.0759 | 0.0215 |
0.1929 | 50.99 | 1963 | 0.0927 | 0.0754 | 0.0212 |
0.183 | 52.0 | 2002 | 0.0927 | 0.0750 | 0.0215 |
0.183 | 52.99 | 2040 | 0.0909 | 0.0745 | 0.0212 |
0.183 | 54.0 | 2079 | 0.0920 | 0.0750 | 0.0220 |
0.1812 | 54.99 | 2117 | 0.0910 | 0.0732 | 0.0206 |
0.1812 | 56.0 | 2156 | 0.0922 | 0.0759 | 0.0216 |
0.1812 | 56.99 | 2194 | 0.0916 | 0.0718 | 0.0210 |
0.158 | 58.0 | 2233 | 0.0907 | 0.0732 | 0.0212 |
0.158 | 58.99 | 2271 | 0.0916 | 0.0750 | 0.0209 |
0.1716 | 60.0 | 2310 | 0.0908 | 0.0736 | 0.0214 |
0.1716 | 60.99 | 2348 | 0.0913 | 0.0727 | 0.0211 |
0.1716 | 62.0 | 2387 | 0.0897 | 0.0741 | 0.0213 |
0.154 | 62.99 | 2425 | 0.0896 | 0.0750 | 0.0208 |
0.154 | 64.0 | 2464 | 0.0906 | 0.0754 | 0.0212 |
0.1598 | 64.99 | 2502 | 0.0892 | 0.0776 | 0.0212 |
0.1598 | 66.0 | 2541 | 0.0907 | 0.0759 | 0.0209 |
0.1598 | 66.99 | 2579 | 0.0894 | 0.0732 | 0.0204 |
0.1484 | 68.0 | 2618 | 0.0918 | 0.0754 | 0.0210 |
0.1484 | 68.99 | 2656 | 0.0892 | 0.0754 | 0.0212 |
0.1484 | 70.0 | 2695 | 0.0897 | 0.0705 | 0.0205 |
0.1841 | 70.99 | 2733 | 0.0885 | 0.0736 | 0.0203 |
0.1841 | 72.0 | 2772 | 0.0892 | 0.0745 | 0.0211 |
0.1629 | 72.99 | 2810 | 0.0887 | 0.0732 | 0.0212 |
0.1629 | 74.0 | 2849 | 0.0893 | 0.0723 | 0.0205 |
0.1629 | 74.99 | 2887 | 0.0877 | 0.0754 | 0.0204 |
0.1549 | 76.0 | 2926 | 0.0877 | 0.0736 | 0.0206 |
0.1549 | 76.99 | 2964 | 0.0862 | 0.0727 | 0.0203 |
0.1642 | 78.0 | 3003 | 0.0878 | 0.0718 | 0.0201 |
0.1642 | 78.99 | 3041 | 0.0869 | 0.0718 | 0.0203 |
0.1642 | 80.0 | 3080 | 0.0867 | 0.0723 | 0.0205 |
0.1356 | 80.99 | 3118 | 0.0879 | 0.0732 | 0.0207 |
0.1356 | 82.0 | 3157 | 0.0872 | 0.0700 | 0.0202 |
0.1356 | 82.99 | 3195 | 0.0867 | 0.0700 | 0.0200 |
0.1376 | 84.0 | 3234 | 0.0871 | 0.0732 | 0.0203 |
0.1376 | 84.99 | 3272 | 0.0873 | 0.0727 | 0.0201 |
0.1381 | 86.0 | 3311 | 0.0878 | 0.0736 | 0.0203 |
0.1381 | 86.99 | 3349 | 0.0855 | 0.0709 | 0.0199 |
0.1381 | 88.0 | 3388 | 0.0861 | 0.0723 | 0.0200 |
0.1497 | 88.99 | 3426 | 0.0853 | 0.0727 | 0.0198 |
0.1497 | 90.0 | 3465 | 0.0854 | 0.0718 | 0.0197 |
0.1326 | 90.99 | 3503 | 0.0855 | 0.0736 | 0.0200 |
0.1326 | 92.0 | 3542 | 0.0849 | 0.0727 | 0.0202 |
0.1326 | 92.99 | 3580 | 0.0848 | 0.0736 | 0.0202 |
0.1506 | 94.0 | 3619 | 0.0849 | 0.0741 | 0.0203 |
0.1506 | 94.99 | 3657 | 0.0851 | 0.0754 | 0.0204 |
0.1506 | 96.0 | 3696 | 0.0855 | 0.0736 | 0.0204 |
0.1288 | 96.99 | 3734 | 0.0855 | 0.0745 | 0.0202 |
0.1288 | 98.0 | 3773 | 0.0853 | 0.0750 | 0.0201 |
0.1307 | 98.7 | 3800 | 0.0855 | 0.0745 | 0.0200 |
Framework versions
- Transformers 4.28.0
- Pytorch 2.0.1+cu117
- Datasets 2.14.4
- Tokenizers 0.13.3