<!-- 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-3-4
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: 0.0974
- Wer: 0.0795
- Cer: 0.0234
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 |
---|---|---|---|---|---|
23.7699 | 0.98 | 28 | 6.1258 | 1.0 | 1.0 |
23.7699 | 2.0 | 57 | 3.4198 | 1.0 | 1.0 |
23.7699 | 2.98 | 85 | 3.0944 | 1.0 | 1.0 |
6.5792 | 4.0 | 114 | 2.9843 | 1.0 | 1.0 |
6.5792 | 4.98 | 142 | 2.9231 | 1.0 | 1.0 |
6.5792 | 6.0 | 171 | 2.9258 | 1.0 | 1.0 |
6.5792 | 6.98 | 199 | 2.8925 | 1.0 | 1.0 |
2.9859 | 8.0 | 228 | 2.8992 | 1.0 | 1.0 |
2.9859 | 8.98 | 256 | 2.8759 | 1.0 | 1.0 |
2.9859 | 10.0 | 285 | 2.8514 | 1.0 | 1.0 |
2.8871 | 10.98 | 313 | 2.7826 | 1.0 | 1.0 |
2.8871 | 12.0 | 342 | 2.5686 | 1.0 | 0.9937 |
2.8871 | 12.98 | 370 | 1.9836 | 1.0 | 0.6463 |
2.8871 | 14.0 | 399 | 1.1686 | 0.9959 | 0.2919 |
2.4534 | 14.98 | 427 | 0.7458 | 0.3377 | 0.0798 |
2.4534 | 16.0 | 456 | 0.5120 | 0.1952 | 0.0494 |
2.4534 | 16.98 | 484 | 0.4184 | 0.1719 | 0.0458 |
1.0591 | 18.0 | 513 | 0.3535 | 0.1808 | 0.0454 |
1.0591 | 18.98 | 541 | 0.3105 | 0.1534 | 0.0399 |
1.0591 | 20.0 | 570 | 0.2834 | 0.1514 | 0.0390 |
1.0591 | 20.98 | 598 | 0.2566 | 0.1466 | 0.0373 |
0.6207 | 22.0 | 627 | 0.2380 | 0.1486 | 0.0382 |
0.6207 | 22.98 | 655 | 0.2218 | 0.1411 | 0.0365 |
0.6207 | 24.0 | 684 | 0.2116 | 0.1438 | 0.0368 |
0.4915 | 24.98 | 712 | 0.1993 | 0.1411 | 0.0352 |
0.4915 | 26.0 | 741 | 0.1906 | 0.1308 | 0.0338 |
0.4915 | 26.98 | 769 | 0.1824 | 0.1205 | 0.0320 |
0.4915 | 28.0 | 798 | 0.1758 | 0.1205 | 0.0312 |
0.3813 | 28.98 | 826 | 0.1754 | 0.1178 | 0.0317 |
0.3813 | 30.0 | 855 | 0.1732 | 0.1096 | 0.0303 |
0.3813 | 30.98 | 883 | 0.1703 | 0.1144 | 0.0310 |
0.3389 | 32.0 | 912 | 0.1647 | 0.1055 | 0.0295 |
0.3389 | 32.98 | 940 | 0.1579 | 0.1 | 0.0275 |
0.3389 | 34.0 | 969 | 0.1531 | 0.1 | 0.0279 |
0.3389 | 34.98 | 997 | 0.1469 | 0.1 | 0.0267 |
0.3337 | 36.0 | 1026 | 0.1448 | 0.1014 | 0.0278 |
0.3337 | 36.98 | 1054 | 0.1423 | 0.0932 | 0.0263 |
0.3337 | 38.0 | 1083 | 0.1383 | 0.0973 | 0.0270 |
0.2783 | 38.98 | 1111 | 0.1354 | 0.0938 | 0.0263 |
0.2783 | 40.0 | 1140 | 0.1312 | 0.0863 | 0.0255 |
0.2783 | 40.98 | 1168 | 0.1318 | 0.0925 | 0.0263 |
0.2783 | 42.0 | 1197 | 0.1274 | 0.0904 | 0.0260 |
0.2517 | 42.98 | 1225 | 0.1289 | 0.0932 | 0.0267 |
0.2517 | 44.0 | 1254 | 0.1258 | 0.0890 | 0.0257 |
0.2517 | 44.98 | 1282 | 0.1226 | 0.0856 | 0.0251 |
0.2478 | 46.0 | 1311 | 0.1219 | 0.0911 | 0.0261 |
0.2478 | 46.98 | 1339 | 0.1205 | 0.0884 | 0.0261 |
0.2478 | 48.0 | 1368 | 0.1221 | 0.0870 | 0.0265 |
0.2478 | 48.98 | 1396 | 0.1178 | 0.0870 | 0.0261 |
0.2323 | 50.0 | 1425 | 0.1143 | 0.0842 | 0.0248 |
0.2323 | 50.98 | 1453 | 0.1136 | 0.0829 | 0.0249 |
0.2323 | 52.0 | 1482 | 0.1145 | 0.0829 | 0.0243 |
0.2166 | 52.98 | 1510 | 0.1151 | 0.0836 | 0.0249 |
0.2166 | 54.0 | 1539 | 0.1135 | 0.0849 | 0.0255 |
0.2166 | 54.98 | 1567 | 0.1145 | 0.0870 | 0.0256 |
0.2166 | 56.0 | 1596 | 0.1096 | 0.0836 | 0.0242 |
0.2184 | 56.98 | 1624 | 0.1103 | 0.0842 | 0.0240 |
0.2184 | 58.0 | 1653 | 0.1104 | 0.0877 | 0.0255 |
0.2184 | 58.98 | 1681 | 0.1110 | 0.0863 | 0.0248 |
0.1997 | 60.0 | 1710 | 0.1109 | 0.0863 | 0.0255 |
0.1997 | 60.98 | 1738 | 0.1106 | 0.0863 | 0.0253 |
0.1997 | 62.0 | 1767 | 0.1100 | 0.0877 | 0.0252 |
0.1997 | 62.98 | 1795 | 0.1117 | 0.0863 | 0.0253 |
0.2004 | 64.0 | 1824 | 0.1098 | 0.0849 | 0.0247 |
0.2004 | 64.98 | 1852 | 0.1087 | 0.0856 | 0.0256 |
0.2004 | 66.0 | 1881 | 0.1072 | 0.0849 | 0.0249 |
0.1959 | 66.98 | 1909 | 0.1074 | 0.0863 | 0.0253 |
0.1959 | 68.0 | 1938 | 0.1080 | 0.0829 | 0.0246 |
0.1959 | 68.98 | 1966 | 0.1057 | 0.0815 | 0.0240 |
0.1959 | 70.0 | 1995 | 0.1057 | 0.0856 | 0.0247 |
0.1849 | 70.98 | 2023 | 0.1055 | 0.0836 | 0.0243 |
0.1849 | 72.0 | 2052 | 0.1058 | 0.0863 | 0.0247 |
0.1849 | 72.98 | 2080 | 0.1052 | 0.0863 | 0.0248 |
0.1687 | 74.0 | 2109 | 0.1043 | 0.0856 | 0.0244 |
0.1687 | 74.98 | 2137 | 0.1025 | 0.0849 | 0.0247 |
0.1687 | 76.0 | 2166 | 0.1026 | 0.0856 | 0.0249 |
0.1687 | 76.98 | 2194 | 0.1022 | 0.0842 | 0.0246 |
0.1701 | 78.0 | 2223 | 0.1016 | 0.0856 | 0.0244 |
0.1701 | 78.98 | 2251 | 0.1009 | 0.0870 | 0.0251 |
0.1701 | 80.0 | 2280 | 0.1015 | 0.0829 | 0.0244 |
0.1739 | 80.98 | 2308 | 0.1021 | 0.0836 | 0.0244 |
0.1739 | 82.0 | 2337 | 0.1023 | 0.0829 | 0.0244 |
0.1739 | 82.98 | 2365 | 0.1015 | 0.0829 | 0.0243 |
0.1739 | 84.0 | 2394 | 0.1016 | 0.0795 | 0.0237 |
0.1565 | 84.98 | 2422 | 0.1005 | 0.0808 | 0.0239 |
0.1565 | 86.0 | 2451 | 0.1011 | 0.0815 | 0.0240 |
0.1565 | 86.98 | 2479 | 0.1007 | 0.0822 | 0.0242 |
0.1695 | 88.0 | 2508 | 0.1000 | 0.0822 | 0.0240 |
0.1695 | 88.98 | 2536 | 0.0994 | 0.0815 | 0.0240 |
0.1695 | 90.0 | 2565 | 0.0990 | 0.0815 | 0.0239 |
0.1695 | 90.98 | 2593 | 0.0986 | 0.0815 | 0.0240 |
0.1579 | 92.0 | 2622 | 0.0985 | 0.0801 | 0.0237 |
0.1579 | 92.98 | 2650 | 0.0988 | 0.0795 | 0.0237 |
0.1579 | 94.0 | 2679 | 0.0979 | 0.0788 | 0.0234 |
0.1632 | 94.98 | 2707 | 0.0976 | 0.0795 | 0.0235 |
0.1632 | 96.0 | 2736 | 0.0982 | 0.0808 | 0.0237 |
0.1632 | 96.98 | 2764 | 0.0979 | 0.0795 | 0.0234 |
0.1632 | 98.0 | 2793 | 0.0974 | 0.0795 | 0.0234 |
0.1736 | 98.25 | 2800 | 0.0980 | 0.0801 | 0.0235 |
Framework versions
- Transformers 4.28.0
- Pytorch 2.0.1+cu117
- Datasets 2.14.4
- Tokenizers 0.13.3