<!-- 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
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: 0.2772
- Wer: 0.1515
- Cer: 0.0437
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 |
---|---|---|---|---|---|
33.3841 | 0.98 | 29 | 7.7518 | 1.0 | 1.0 |
33.3841 | 2.0 | 59 | 3.6287 | 1.0 | 1.0 |
33.3841 | 2.98 | 88 | 3.2594 | 1.0 | 1.0 |
9.6274 | 4.0 | 118 | 3.0839 | 1.0 | 1.0 |
9.6274 | 4.98 | 147 | 2.9796 | 1.0 | 1.0 |
9.6274 | 6.0 | 177 | 2.9473 | 1.0 | 1.0 |
3.0658 | 6.98 | 206 | 2.9286 | 1.0 | 1.0 |
3.0658 | 8.0 | 236 | 2.9359 | 1.0 | 1.0 |
3.0658 | 8.98 | 265 | 2.9138 | 1.0 | 1.0 |
3.0658 | 10.0 | 295 | 2.9180 | 1.0 | 1.0 |
2.9458 | 10.98 | 324 | 2.9141 | 1.0 | 1.0 |
2.9458 | 12.0 | 354 | 2.8907 | 1.0 | 1.0 |
2.9458 | 12.98 | 383 | 2.8719 | 1.0 | 1.0 |
2.9146 | 14.0 | 413 | 2.8021 | 1.0 | 1.0 |
2.9146 | 14.98 | 442 | 2.7561 | 1.0 | 1.0 |
2.9146 | 16.0 | 472 | 2.5043 | 0.9977 | 0.8479 |
2.6953 | 16.98 | 501 | 1.9270 | 1.0 | 0.6237 |
2.6953 | 18.0 | 531 | 1.2892 | 1.0 | 0.3074 |
2.6953 | 18.98 | 560 | 0.9609 | 0.9336 | 0.2278 |
2.6953 | 20.0 | 590 | 0.7426 | 0.3205 | 0.0829 |
1.496 | 20.98 | 619 | 0.6321 | 0.2552 | 0.0701 |
1.496 | 22.0 | 649 | 0.5499 | 0.2401 | 0.0666 |
1.496 | 22.98 | 678 | 0.4959 | 0.2273 | 0.0628 |
0.7943 | 24.0 | 708 | 0.4692 | 0.2284 | 0.0617 |
0.7943 | 24.98 | 737 | 0.4361 | 0.2191 | 0.0603 |
0.7943 | 26.0 | 767 | 0.4278 | 0.2110 | 0.0596 |
0.7943 | 26.98 | 796 | 0.4133 | 0.2110 | 0.0591 |
0.5809 | 28.0 | 826 | 0.3929 | 0.2028 | 0.0579 |
0.5809 | 28.98 | 855 | 0.3908 | 0.2086 | 0.0589 |
0.5809 | 30.0 | 885 | 0.3825 | 0.1946 | 0.0563 |
0.4752 | 30.98 | 914 | 0.3759 | 0.1876 | 0.0547 |
0.4752 | 32.0 | 944 | 0.3610 | 0.1841 | 0.0535 |
0.4752 | 32.98 | 973 | 0.3741 | 0.1783 | 0.0516 |
0.3928 | 34.0 | 1003 | 0.3499 | 0.1772 | 0.0512 |
0.3928 | 34.98 | 1032 | 0.3406 | 0.1655 | 0.0488 |
0.3928 | 36.0 | 1062 | 0.3470 | 0.1678 | 0.0495 |
0.3928 | 36.98 | 1091 | 0.3335 | 0.1678 | 0.0486 |
0.3676 | 38.0 | 1121 | 0.3452 | 0.1725 | 0.0502 |
0.3676 | 38.98 | 1150 | 0.3433 | 0.1643 | 0.0498 |
0.3676 | 40.0 | 1180 | 0.3314 | 0.1655 | 0.0491 |
0.3164 | 40.98 | 1209 | 0.3392 | 0.1620 | 0.0493 |
0.3164 | 42.0 | 1239 | 0.3238 | 0.1608 | 0.0484 |
0.3164 | 42.98 | 1268 | 0.3214 | 0.1597 | 0.0477 |
0.3164 | 44.0 | 1298 | 0.3302 | 0.1632 | 0.0491 |
0.2886 | 44.98 | 1327 | 0.3298 | 0.1585 | 0.0486 |
0.2886 | 46.0 | 1357 | 0.3187 | 0.1573 | 0.0472 |
0.2886 | 46.98 | 1386 | 0.3039 | 0.1632 | 0.0472 |
0.2835 | 48.0 | 1416 | 0.3098 | 0.1597 | 0.0460 |
0.2835 | 48.98 | 1445 | 0.3144 | 0.1585 | 0.0465 |
0.2835 | 50.0 | 1475 | 0.3045 | 0.1527 | 0.0453 |
0.2325 | 50.98 | 1504 | 0.3092 | 0.1480 | 0.0448 |
0.2325 | 52.0 | 1534 | 0.3217 | 0.1608 | 0.0479 |
0.2325 | 52.98 | 1563 | 0.3039 | 0.1562 | 0.0460 |
0.2325 | 54.0 | 1593 | 0.2991 | 0.1527 | 0.0446 |
0.2445 | 54.98 | 1622 | 0.2948 | 0.1515 | 0.0446 |
0.2445 | 56.0 | 1652 | 0.3062 | 0.1550 | 0.0460 |
0.2445 | 56.98 | 1681 | 0.3119 | 0.1515 | 0.0446 |
0.2388 | 58.0 | 1711 | 0.2998 | 0.1515 | 0.0444 |
0.2388 | 58.98 | 1740 | 0.2984 | 0.1550 | 0.0441 |
0.2388 | 60.0 | 1770 | 0.2936 | 0.1527 | 0.0441 |
0.2388 | 60.98 | 1799 | 0.2958 | 0.1527 | 0.0448 |
0.2139 | 62.0 | 1829 | 0.2922 | 0.1469 | 0.0434 |
0.2139 | 62.98 | 1858 | 0.2932 | 0.1503 | 0.0441 |
0.2139 | 64.0 | 1888 | 0.2970 | 0.1550 | 0.0446 |
0.2209 | 64.98 | 1917 | 0.2993 | 0.1503 | 0.0453 |
0.2209 | 66.0 | 1947 | 0.2990 | 0.1480 | 0.0441 |
0.2209 | 66.98 | 1976 | 0.2966 | 0.1480 | 0.0444 |
0.2039 | 68.0 | 2006 | 0.2910 | 0.1503 | 0.0444 |
0.2039 | 68.98 | 2035 | 0.2834 | 0.1469 | 0.0437 |
0.2039 | 70.0 | 2065 | 0.2843 | 0.1480 | 0.0437 |
0.2039 | 70.98 | 2094 | 0.2784 | 0.1550 | 0.0437 |
0.1907 | 72.0 | 2124 | 0.2848 | 0.1515 | 0.0437 |
0.1907 | 72.98 | 2153 | 0.2845 | 0.1515 | 0.0439 |
0.1907 | 74.0 | 2183 | 0.2855 | 0.1457 | 0.0441 |
0.1951 | 74.98 | 2212 | 0.2878 | 0.1492 | 0.0430 |
0.1951 | 76.0 | 2242 | 0.2888 | 0.1469 | 0.0432 |
0.1951 | 76.98 | 2271 | 0.2847 | 0.1492 | 0.0441 |
0.185 | 78.0 | 2301 | 0.2847 | 0.1469 | 0.0437 |
0.185 | 78.98 | 2330 | 0.2818 | 0.1457 | 0.0423 |
0.185 | 80.0 | 2360 | 0.2822 | 0.1457 | 0.0432 |
0.185 | 80.98 | 2389 | 0.2805 | 0.1480 | 0.0434 |
0.1964 | 82.0 | 2419 | 0.2828 | 0.1492 | 0.0430 |
0.1964 | 82.98 | 2448 | 0.2772 | 0.1515 | 0.0437 |
0.1964 | 84.0 | 2478 | 0.2830 | 0.1434 | 0.0430 |
0.1816 | 84.98 | 2507 | 0.2809 | 0.1492 | 0.0437 |
0.1816 | 86.0 | 2537 | 0.2810 | 0.1469 | 0.0432 |
0.1816 | 86.98 | 2566 | 0.2838 | 0.1469 | 0.0432 |
0.1816 | 88.0 | 2596 | 0.2806 | 0.1457 | 0.0425 |
0.1819 | 88.98 | 2625 | 0.2832 | 0.1445 | 0.0432 |
0.1819 | 90.0 | 2655 | 0.2844 | 0.1469 | 0.0430 |
0.1819 | 90.98 | 2684 | 0.2840 | 0.1469 | 0.0425 |
0.1841 | 92.0 | 2714 | 0.2813 | 0.1480 | 0.0432 |
0.1841 | 92.98 | 2743 | 0.2833 | 0.1492 | 0.0437 |
0.1841 | 94.0 | 2773 | 0.2824 | 0.1480 | 0.0437 |
0.17 | 94.98 | 2802 | 0.2822 | 0.1480 | 0.0432 |
0.17 | 96.0 | 2832 | 0.2813 | 0.1480 | 0.0432 |
0.17 | 96.98 | 2861 | 0.2830 | 0.1480 | 0.0427 |
0.17 | 98.0 | 2891 | 0.2829 | 0.1503 | 0.0432 |
0.1731 | 98.31 | 2900 | 0.2827 | 0.1469 | 0.0432 |
Framework versions
- Transformers 4.28.0
- Pytorch 2.0.1+cu117
- Datasets 2.14.4
- Tokenizers 0.13.3