<!-- 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-2-4-5
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: 0.1692
- Wer: 0.0918
- Cer: 0.0281
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 |
---|---|---|---|---|---|
28.5301 | 1.0 | 65 | 9.1162 | 1.0 | 0.9852 |
12.9321 | 2.0 | 130 | 6.5974 | 1.0 | 0.9746 |
12.9321 | 3.0 | 195 | 3.1146 | 1.0 | 1.0 |
5.5737 | 4.0 | 260 | 2.9780 | 1.0 | 1.0 |
3.0199 | 5.0 | 325 | 2.9426 | 1.0 | 1.0 |
3.0199 | 6.0 | 390 | 2.9347 | 1.0 | 1.0 |
2.9212 | 7.0 | 455 | 2.8947 | 1.0 | 1.0 |
2.9 | 8.0 | 520 | 2.8812 | 1.0 | 1.0 |
2.9 | 9.0 | 585 | 2.5546 | 1.0 | 1.0 |
2.792 | 10.0 | 650 | 1.2976 | 0.9997 | 0.3312 |
1.769 | 11.0 | 715 | 0.6302 | 0.3400 | 0.0873 |
1.769 | 12.0 | 780 | 0.4592 | 0.2271 | 0.0621 |
0.8629 | 13.0 | 845 | 0.3916 | 0.2025 | 0.0566 |
0.6402 | 14.0 | 910 | 0.3392 | 0.1775 | 0.0496 |
0.6402 | 15.0 | 975 | 0.3035 | 0.1652 | 0.0458 |
0.5193 | 16.0 | 1040 | 0.2981 | 0.1674 | 0.0464 |
0.4874 | 17.0 | 1105 | 0.2786 | 0.1528 | 0.0435 |
0.4874 | 18.0 | 1170 | 0.2596 | 0.1434 | 0.0409 |
0.424 | 19.0 | 1235 | 0.2530 | 0.1347 | 0.0389 |
0.3769 | 20.0 | 1300 | 0.2429 | 0.1252 | 0.0370 |
0.3769 | 21.0 | 1365 | 0.2323 | 0.1278 | 0.0370 |
0.3467 | 22.0 | 1430 | 0.2281 | 0.1136 | 0.0346 |
0.3467 | 23.0 | 1495 | 0.2177 | 0.1090 | 0.0331 |
0.3269 | 24.0 | 1560 | 0.2193 | 0.1090 | 0.0332 |
0.3055 | 25.0 | 1625 | 0.2155 | 0.1071 | 0.0320 |
0.3055 | 26.0 | 1690 | 0.2035 | 0.1012 | 0.0315 |
0.2932 | 27.0 | 1755 | 0.2121 | 0.0993 | 0.0310 |
0.2951 | 28.0 | 1820 | 0.2101 | 0.1025 | 0.0311 |
0.2951 | 29.0 | 1885 | 0.2035 | 0.1025 | 0.0314 |
0.2677 | 30.0 | 1950 | 0.1990 | 0.1025 | 0.0313 |
0.2624 | 31.0 | 2015 | 0.1934 | 0.0980 | 0.0307 |
0.2624 | 32.0 | 2080 | 0.1989 | 0.0957 | 0.0298 |
0.2536 | 33.0 | 2145 | 0.1941 | 0.0960 | 0.0298 |
0.2415 | 34.0 | 2210 | 0.1904 | 0.1019 | 0.0305 |
0.2415 | 35.0 | 2275 | 0.1817 | 0.0954 | 0.0288 |
0.22 | 36.0 | 2340 | 0.1835 | 0.0934 | 0.0288 |
0.2358 | 37.0 | 2405 | 0.1808 | 0.0912 | 0.0286 |
0.2358 | 38.0 | 2470 | 0.1798 | 0.0931 | 0.0285 |
0.2152 | 39.0 | 2535 | 0.1815 | 0.0938 | 0.0286 |
0.2031 | 40.0 | 2600 | 0.1801 | 0.0960 | 0.0287 |
0.2031 | 41.0 | 2665 | 0.1800 | 0.0909 | 0.0286 |
0.2144 | 42.0 | 2730 | 0.1781 | 0.0928 | 0.0287 |
0.2144 | 43.0 | 2795 | 0.1801 | 0.0886 | 0.0283 |
0.2038 | 44.0 | 2860 | 0.1806 | 0.0921 | 0.0280 |
0.1992 | 45.0 | 2925 | 0.1784 | 0.0915 | 0.0283 |
0.1992 | 46.0 | 2990 | 0.1863 | 0.0957 | 0.0288 |
0.18 | 47.0 | 3055 | 0.1786 | 0.0915 | 0.0280 |
0.1796 | 48.0 | 3120 | 0.1800 | 0.0896 | 0.0281 |
0.1796 | 49.0 | 3185 | 0.1781 | 0.0915 | 0.0285 |
0.1914 | 50.0 | 3250 | 0.1783 | 0.0883 | 0.0278 |
0.1875 | 51.0 | 3315 | 0.1791 | 0.0918 | 0.0282 |
0.1875 | 52.0 | 3380 | 0.1758 | 0.0899 | 0.0282 |
0.1857 | 53.0 | 3445 | 0.1771 | 0.0883 | 0.0274 |
0.2022 | 54.0 | 3510 | 0.1808 | 0.0902 | 0.0280 |
0.2022 | 55.0 | 3575 | 0.1754 | 0.0879 | 0.0268 |
0.1697 | 56.0 | 3640 | 0.1724 | 0.0879 | 0.0268 |
0.1663 | 57.0 | 3705 | 0.1719 | 0.0870 | 0.0271 |
0.1663 | 58.0 | 3770 | 0.1692 | 0.0918 | 0.0281 |
0.1627 | 59.0 | 3835 | 0.1710 | 0.0899 | 0.0273 |
0.174 | 60.0 | 3900 | 0.1745 | 0.0863 | 0.0270 |
0.174 | 61.0 | 3965 | 0.1757 | 0.0938 | 0.0285 |
0.1665 | 62.0 | 4030 | 0.1755 | 0.0921 | 0.0279 |
0.1665 | 63.0 | 4095 | 0.1805 | 0.0915 | 0.0283 |
0.154 | 64.0 | 4160 | 0.1820 | 0.0883 | 0.0282 |
0.1508 | 65.0 | 4225 | 0.1809 | 0.0870 | 0.0280 |
0.1508 | 66.0 | 4290 | 0.1823 | 0.0883 | 0.0285 |
0.1601 | 67.0 | 4355 | 0.1811 | 0.0915 | 0.0285 |
0.1511 | 68.0 | 4420 | 0.1785 | 0.0886 | 0.0279 |
0.1511 | 69.0 | 4485 | 0.1745 | 0.0876 | 0.0279 |
0.1598 | 70.0 | 4550 | 0.1766 | 0.0879 | 0.0280 |
0.1606 | 71.0 | 4615 | 0.1765 | 0.0883 | 0.0279 |
0.1606 | 72.0 | 4680 | 0.1783 | 0.0883 | 0.0280 |
0.1523 | 73.0 | 4745 | 0.1788 | 0.0873 | 0.0278 |
0.1523 | 74.0 | 4810 | 0.1777 | 0.0886 | 0.0277 |
0.1523 | 75.0 | 4875 | 0.1809 | 0.0870 | 0.0277 |
0.1434 | 76.0 | 4940 | 0.1810 | 0.0921 | 0.0287 |
0.1428 | 77.0 | 5005 | 0.1821 | 0.0886 | 0.0285 |
0.1428 | 78.0 | 5070 | 0.1792 | 0.0860 | 0.0275 |
Framework versions
- Transformers 4.28.0
- Pytorch 2.0.1+cu117
- Datasets 2.14.4
- Tokenizers 0.13.3