<!-- 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-clean-grade-2-3-4
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: 0.1674
- Wer: 0.0873
- Cer: 0.0263
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 |
---|---|---|---|---|---|
29.5686 | 1.0 | 51 | 3.7627 | 1.0 | 1.0 |
8.7555 | 2.0 | 102 | 3.1584 | 1.0 | 1.0 |
8.7555 | 3.0 | 153 | 2.9588 | 1.0 | 1.0 |
3.0494 | 4.0 | 204 | 2.9116 | 1.0 | 1.0 |
3.0494 | 5.0 | 255 | 2.8915 | 1.0 | 1.0 |
2.9145 | 6.0 | 306 | 2.8145 | 1.0 | 1.0 |
2.9145 | 7.0 | 357 | 2.4510 | 0.9972 | 0.9959 |
2.5483 | 8.0 | 408 | 1.2676 | 0.9963 | 0.2686 |
2.5483 | 9.0 | 459 | 0.6644 | 0.3279 | 0.0786 |
1.244 | 10.0 | 510 | 0.4906 | 0.2069 | 0.0549 |
1.244 | 11.0 | 561 | 0.4020 | 0.1752 | 0.0467 |
0.6905 | 12.0 | 612 | 0.3483 | 0.1546 | 0.0440 |
0.6905 | 13.0 | 663 | 0.3099 | 0.1532 | 0.0418 |
0.5135 | 14.0 | 714 | 0.2802 | 0.1383 | 0.0383 |
0.5135 | 15.0 | 765 | 0.2640 | 0.1312 | 0.0366 |
0.4224 | 16.0 | 816 | 0.2474 | 0.1228 | 0.0330 |
0.4224 | 17.0 | 867 | 0.2358 | 0.1252 | 0.0332 |
0.3686 | 18.0 | 918 | 0.2304 | 0.1121 | 0.0311 |
0.3686 | 19.0 | 969 | 0.2224 | 0.1126 | 0.0323 |
0.3442 | 20.0 | 1020 | 0.2213 | 0.1028 | 0.0301 |
0.3442 | 21.0 | 1071 | 0.2084 | 0.1014 | 0.0292 |
0.3164 | 22.0 | 1122 | 0.2125 | 0.1060 | 0.0300 |
0.3164 | 23.0 | 1173 | 0.2011 | 0.1046 | 0.0293 |
0.2821 | 24.0 | 1224 | 0.1996 | 0.0929 | 0.0278 |
0.2821 | 25.0 | 1275 | 0.1923 | 0.0929 | 0.0275 |
0.2638 | 26.0 | 1326 | 0.1966 | 0.0957 | 0.0285 |
0.2638 | 27.0 | 1377 | 0.1942 | 0.0906 | 0.0277 |
0.2438 | 28.0 | 1428 | 0.1866 | 0.0934 | 0.0272 |
0.2438 | 29.0 | 1479 | 0.1899 | 0.0920 | 0.0275 |
0.2438 | 30.0 | 1530 | 0.1805 | 0.0878 | 0.0260 |
0.2438 | 31.0 | 1581 | 0.1760 | 0.0920 | 0.0270 |
0.2185 | 32.0 | 1632 | 0.1850 | 0.0887 | 0.0263 |
0.2185 | 33.0 | 1683 | 0.1816 | 0.0869 | 0.0263 |
0.2097 | 34.0 | 1734 | 0.1823 | 0.0897 | 0.0259 |
0.2097 | 35.0 | 1785 | 0.1780 | 0.0883 | 0.0258 |
0.1909 | 36.0 | 1836 | 0.1813 | 0.0897 | 0.0259 |
0.1909 | 37.0 | 1887 | 0.1776 | 0.0883 | 0.0258 |
0.1896 | 38.0 | 1938 | 0.1733 | 0.0855 | 0.0259 |
0.1896 | 39.0 | 1989 | 0.1698 | 0.0864 | 0.0262 |
0.2024 | 40.0 | 2040 | 0.1729 | 0.0897 | 0.0265 |
0.2024 | 41.0 | 2091 | 0.1799 | 0.0873 | 0.0266 |
0.1846 | 42.0 | 2142 | 0.1817 | 0.0901 | 0.0263 |
0.1846 | 43.0 | 2193 | 0.1810 | 0.0859 | 0.0263 |
0.1839 | 44.0 | 2244 | 0.1801 | 0.0887 | 0.0264 |
0.1839 | 45.0 | 2295 | 0.1802 | 0.0869 | 0.0259 |
0.1674 | 46.0 | 2346 | 0.1774 | 0.0967 | 0.0265 |
0.1674 | 47.0 | 2397 | 0.1768 | 0.0883 | 0.0258 |
0.1689 | 48.0 | 2448 | 0.1774 | 0.0887 | 0.0257 |
0.1689 | 49.0 | 2499 | 0.1750 | 0.0859 | 0.0253 |
0.1641 | 50.0 | 2550 | 0.1780 | 0.0925 | 0.0266 |
0.1566 | 51.0 | 2601 | 0.1726 | 0.0878 | 0.0256 |
0.1566 | 52.0 | 2652 | 0.1749 | 0.0873 | 0.0264 |
0.1629 | 53.0 | 2703 | 0.1684 | 0.0878 | 0.0257 |
0.1629 | 54.0 | 2754 | 0.1726 | 0.0864 | 0.0255 |
0.15 | 55.0 | 2805 | 0.1727 | 0.0850 | 0.0250 |
0.15 | 56.0 | 2856 | 0.1747 | 0.0864 | 0.0256 |
0.1541 | 57.0 | 2907 | 0.1729 | 0.0845 | 0.0249 |
0.1541 | 58.0 | 2958 | 0.1705 | 0.0827 | 0.0255 |
0.1456 | 59.0 | 3009 | 0.1681 | 0.0859 | 0.0252 |
0.1456 | 60.0 | 3060 | 0.1730 | 0.0883 | 0.0258 |
0.1395 | 61.0 | 3111 | 0.1721 | 0.0836 | 0.0250 |
0.1395 | 62.0 | 3162 | 0.1707 | 0.0892 | 0.0257 |
0.1336 | 63.0 | 3213 | 0.1745 | 0.0892 | 0.0258 |
0.1336 | 64.0 | 3264 | 0.1764 | 0.0873 | 0.0259 |
0.1403 | 65.0 | 3315 | 0.1726 | 0.0873 | 0.0255 |
0.1403 | 66.0 | 3366 | 0.1777 | 0.0887 | 0.0260 |
0.1421 | 67.0 | 3417 | 0.1745 | 0.0920 | 0.0264 |
0.1421 | 68.0 | 3468 | 0.1749 | 0.0897 | 0.0266 |
0.1252 | 69.0 | 3519 | 0.1762 | 0.0929 | 0.0271 |
0.1252 | 70.0 | 3570 | 0.1736 | 0.0929 | 0.0267 |
0.137 | 71.0 | 3621 | 0.1744 | 0.0906 | 0.0264 |
0.137 | 72.0 | 3672 | 0.1695 | 0.0883 | 0.0265 |
0.1227 | 73.0 | 3723 | 0.1715 | 0.0883 | 0.0260 |
0.1227 | 74.0 | 3774 | 0.1682 | 0.0920 | 0.0265 |
0.1403 | 75.0 | 3825 | 0.1691 | 0.0836 | 0.0256 |
0.1403 | 76.0 | 3876 | 0.1699 | 0.0883 | 0.0258 |
0.12 | 77.0 | 3927 | 0.1674 | 0.0873 | 0.0263 |
0.12 | 78.0 | 3978 | 0.1700 | 0.0883 | 0.0261 |
0.1275 | 79.0 | 4029 | 0.1691 | 0.0892 | 0.0257 |
0.1275 | 80.0 | 4080 | 0.1698 | 0.0892 | 0.0259 |
0.1313 | 81.0 | 4131 | 0.1719 | 0.0859 | 0.0254 |
0.1313 | 82.0 | 4182 | 0.1758 | 0.0878 | 0.0260 |
0.115 | 83.0 | 4233 | 0.1714 | 0.0873 | 0.0258 |
0.115 | 84.0 | 4284 | 0.1747 | 0.0864 | 0.0253 |
0.1233 | 85.0 | 4335 | 0.1703 | 0.0850 | 0.0250 |
0.1233 | 86.0 | 4386 | 0.1729 | 0.0873 | 0.0257 |
0.1268 | 87.0 | 4437 | 0.1712 | 0.0869 | 0.0253 |
0.1268 | 88.0 | 4488 | 0.1709 | 0.0864 | 0.0251 |
0.1162 | 89.0 | 4539 | 0.1712 | 0.0878 | 0.0255 |
0.1162 | 90.0 | 4590 | 0.1723 | 0.0892 | 0.0254 |
0.118 | 91.0 | 4641 | 0.1723 | 0.0897 | 0.0257 |
0.118 | 92.0 | 4692 | 0.1718 | 0.0892 | 0.0260 |
0.1327 | 93.0 | 4743 | 0.1712 | 0.0887 | 0.0254 |
0.1327 | 94.0 | 4794 | 0.1735 | 0.0892 | 0.0256 |
0.1135 | 95.0 | 4845 | 0.1735 | 0.0892 | 0.0258 |
0.1135 | 96.0 | 4896 | 0.1740 | 0.0883 | 0.0257 |
0.1234 | 97.0 | 4947 | 0.1734 | 0.0883 | 0.0255 |
Framework versions
- Transformers 4.28.0
- Pytorch 2.0.1+cu117
- Datasets 2.14.5
- Tokenizers 0.13.3