<!-- 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-03
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: 0.1367
- Wer: 0.0891
- Cer: 0.0248
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 |
---|---|---|---|---|---|
22.0316 | 1.0 | 67 | 5.8843 | 0.9702 | 0.9853 |
9.9692 | 2.0 | 134 | 4.7064 | 0.9589 | 0.9111 |
5.1963 | 3.0 | 201 | 4.3124 | 0.9609 | 0.9444 |
5.1963 | 4.0 | 268 | 4.1419 | 0.9596 | 0.9452 |
4.5756 | 5.0 | 335 | 4.0705 | 0.9765 | 0.9826 |
4.5148 | 6.0 | 402 | 3.8180 | 0.9755 | 0.9834 |
4.5148 | 7.0 | 469 | 3.8169 | 0.9748 | 0.9835 |
4.1354 | 8.0 | 536 | 3.8965 | 0.9742 | 0.9838 |
3.8029 | 9.0 | 603 | 3.4946 | 0.9752 | 0.9826 |
3.8029 | 10.0 | 670 | 3.3564 | 0.9722 | 0.9834 |
3.6281 | 11.0 | 737 | 3.3778 | 0.9778 | 0.9799 |
3.4586 | 12.0 | 804 | 3.5996 | 0.9735 | 0.9839 |
3.4586 | 13.0 | 871 | 3.0914 | 0.9725 | 0.9843 |
3.2715 | 14.0 | 938 | 3.1400 | 0.9758 | 0.9827 |
3.3073 | 15.0 | 1005 | 3.1388 | 0.9768 | 0.9808 |
3.3073 | 16.0 | 1072 | 3.1733 | 0.9758 | 0.9811 |
3.2006 | 17.0 | 1139 | 3.0084 | 0.9891 | 0.9914 |
3.1493 | 18.0 | 1206 | 3.0215 | 0.9811 | 0.9827 |
3.1493 | 19.0 | 1273 | 3.1791 | 0.9768 | 0.9778 |
3.0254 | 20.0 | 1340 | 3.2284 | 0.9791 | 0.9739 |
2.9615 | 21.0 | 1407 | 3.7736 | 0.9728 | 0.9824 |
2.9615 | 22.0 | 1474 | 3.6841 | 0.9725 | 0.9700 |
2.9588 | 23.0 | 1541 | 3.2069 | 0.9821 | 0.9813 |
2.9151 | 24.0 | 1608 | 2.9156 | 0.9983 | 0.9988 |
2.9151 | 25.0 | 1675 | 2.9429 | 0.9917 | 0.9916 |
2.9081 | 26.0 | 1742 | 3.7887 | 0.9719 | 0.9715 |
2.9139 | 27.0 | 1809 | 2.8008 | 0.9841 | 0.9725 |
2.9139 | 28.0 | 1876 | 2.8962 | 0.9864 | 0.9253 |
2.8165 | 29.0 | 1943 | 2.4966 | 0.9907 | 0.8956 |
2.5363 | 30.0 | 2010 | 1.7632 | 0.9685 | 0.5402 |
2.5363 | 31.0 | 2077 | 0.8292 | 0.5411 | 0.1377 |
1.6241 | 32.0 | 2144 | 0.4653 | 0.3255 | 0.0787 |
0.8674 | 33.0 | 2211 | 0.3569 | 0.2513 | 0.0604 |
0.8674 | 34.0 | 2278 | 0.2987 | 0.1917 | 0.0497 |
0.6012 | 35.0 | 2345 | 0.2660 | 0.1864 | 0.0475 |
0.5177 | 36.0 | 2412 | 0.2435 | 0.1669 | 0.0434 |
0.5177 | 37.0 | 2479 | 0.2229 | 0.1430 | 0.0383 |
0.4461 | 38.0 | 2546 | 0.2097 | 0.1474 | 0.0377 |
0.3942 | 39.0 | 2613 | 0.2031 | 0.1338 | 0.0355 |
0.3942 | 40.0 | 2680 | 0.1968 | 0.1258 | 0.0350 |
0.3685 | 41.0 | 2747 | 0.1909 | 0.1295 | 0.0342 |
0.3445 | 42.0 | 2814 | 0.1795 | 0.1185 | 0.0327 |
0.3445 | 43.0 | 2881 | 0.1807 | 0.1212 | 0.0325 |
0.3284 | 44.0 | 2948 | 0.1740 | 0.1205 | 0.0321 |
0.3117 | 45.0 | 3015 | 0.1686 | 0.1185 | 0.0316 |
0.3117 | 46.0 | 3082 | 0.1693 | 0.1126 | 0.0304 |
0.2911 | 47.0 | 3149 | 0.1656 | 0.1159 | 0.0312 |
0.2866 | 48.0 | 3216 | 0.1619 | 0.1169 | 0.0309 |
0.2866 | 49.0 | 3283 | 0.1638 | 0.1159 | 0.0306 |
0.266 | 50.0 | 3350 | 0.1609 | 0.1083 | 0.0297 |
0.2591 | 51.0 | 3417 | 0.1589 | 0.1083 | 0.0298 |
0.2591 | 52.0 | 3484 | 0.1548 | 0.1043 | 0.0284 |
0.2447 | 53.0 | 3551 | 0.1557 | 0.1070 | 0.0284 |
0.2596 | 54.0 | 3618 | 0.1544 | 0.1033 | 0.0275 |
0.2596 | 55.0 | 3685 | 0.1539 | 0.1060 | 0.0286 |
0.2193 | 56.0 | 3752 | 0.1512 | 0.0977 | 0.0266 |
0.222 | 57.0 | 3819 | 0.1507 | 0.1023 | 0.0276 |
0.222 | 58.0 | 3886 | 0.1478 | 0.1026 | 0.0275 |
0.2266 | 59.0 | 3953 | 0.1461 | 0.0974 | 0.0271 |
0.2168 | 60.0 | 4020 | 0.1447 | 0.0950 | 0.0265 |
0.2168 | 61.0 | 4087 | 0.1458 | 0.0997 | 0.0271 |
0.2057 | 62.0 | 4154 | 0.1470 | 0.0957 | 0.0270 |
0.2039 | 63.0 | 4221 | 0.1450 | 0.0987 | 0.0270 |
0.2039 | 64.0 | 4288 | 0.1429 | 0.0927 | 0.0258 |
0.2049 | 65.0 | 4355 | 0.1422 | 0.1003 | 0.0273 |
0.222 | 66.0 | 4422 | 0.1426 | 0.0997 | 0.0274 |
0.222 | 67.0 | 4489 | 0.1420 | 0.0950 | 0.0261 |
0.2024 | 68.0 | 4556 | 0.1417 | 0.0964 | 0.0265 |
0.2012 | 69.0 | 4623 | 0.1418 | 0.0947 | 0.0261 |
0.2012 | 70.0 | 4690 | 0.1387 | 0.0944 | 0.0260 |
0.1833 | 71.0 | 4757 | 0.1380 | 0.0974 | 0.0267 |
0.1908 | 72.0 | 4824 | 0.1383 | 0.0924 | 0.0259 |
0.1908 | 73.0 | 4891 | 0.1373 | 0.0891 | 0.0253 |
0.1773 | 74.0 | 4958 | 0.1370 | 0.0927 | 0.0257 |
0.1786 | 75.0 | 5025 | 0.1367 | 0.0891 | 0.0248 |
0.1786 | 76.0 | 5092 | 0.1381 | 0.0937 | 0.0260 |
0.1844 | 77.0 | 5159 | 0.1380 | 0.0907 | 0.0251 |
0.1643 | 78.0 | 5226 | 0.1380 | 0.0921 | 0.0253 |
0.1643 | 79.0 | 5293 | 0.1411 | 0.0887 | 0.0253 |
0.1786 | 80.0 | 5360 | 0.1418 | 0.0887 | 0.0253 |
0.1816 | 81.0 | 5427 | 0.1412 | 0.0891 | 0.0253 |
0.1816 | 82.0 | 5494 | 0.1404 | 0.0927 | 0.0256 |
0.1819 | 83.0 | 5561 | 0.1385 | 0.0887 | 0.0249 |
0.1787 | 84.0 | 5628 | 0.1393 | 0.0891 | 0.0247 |
0.1787 | 85.0 | 5695 | 0.1394 | 0.0897 | 0.0253 |
0.1762 | 86.0 | 5762 | 0.1386 | 0.0901 | 0.0254 |
0.1608 | 87.0 | 5829 | 0.1386 | 0.0901 | 0.0255 |
0.1608 | 88.0 | 5896 | 0.1386 | 0.0894 | 0.0252 |
0.1751 | 89.0 | 5963 | 0.1398 | 0.0894 | 0.0253 |
0.1696 | 90.0 | 6030 | 0.1388 | 0.0897 | 0.0251 |
0.1696 | 91.0 | 6097 | 0.1380 | 0.0911 | 0.0253 |
0.1816 | 92.0 | 6164 | 0.1378 | 0.0901 | 0.0253 |
0.1731 | 93.0 | 6231 | 0.1386 | 0.0921 | 0.0255 |
0.1731 | 94.0 | 6298 | 0.1377 | 0.0921 | 0.0255 |
0.1794 | 95.0 | 6365 | 0.1378 | 0.0921 | 0.0254 |
Framework versions
- Transformers 4.28.0
- Pytorch 2.0.0+cu117
- Datasets 2.12.0
- Tokenizers 0.13.3