<!-- 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
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: 0.2273
- Wer: 0.1171
- Cer: 0.0354
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 |
---|---|---|---|---|---|
34.8275 | 0.99 | 40 | 4.4881 | 1.0 | 1.0 |
34.8275 | 2.0 | 81 | 3.3599 | 1.0 | 1.0 |
9.183 | 2.99 | 121 | 3.0961 | 1.0 | 1.0 |
9.183 | 4.0 | 162 | 2.9925 | 1.0 | 1.0 |
3.0834 | 4.99 | 202 | 2.9674 | 1.0 | 1.0 |
3.0834 | 6.0 | 243 | 2.9561 | 1.0 | 1.0 |
3.0834 | 6.99 | 283 | 2.9590 | 1.0 | 1.0 |
2.9575 | 8.0 | 324 | 2.9280 | 1.0 | 1.0 |
2.9575 | 8.99 | 364 | 2.9277 | 1.0 | 1.0 |
2.9217 | 10.0 | 405 | 2.9176 | 1.0 | 1.0 |
2.9217 | 10.99 | 445 | 2.7844 | 1.0 | 1.0 |
2.9217 | 12.0 | 486 | 2.5648 | 1.0 | 0.9999 |
2.7805 | 12.99 | 526 | 1.8401 | 1.0 | 0.5174 |
2.7805 | 14.0 | 567 | 1.0915 | 1.0 | 0.2688 |
1.7493 | 14.99 | 607 | 0.7938 | 0.8164 | 0.1777 |
1.7493 | 16.0 | 648 | 0.5711 | 0.2486 | 0.0689 |
1.7493 | 16.99 | 688 | 0.4946 | 0.235 | 0.0675 |
0.8853 | 18.0 | 729 | 0.4572 | 0.2179 | 0.0644 |
0.8853 | 18.99 | 769 | 0.4036 | 0.2007 | 0.0579 |
0.6608 | 20.0 | 810 | 0.3863 | 0.1993 | 0.0574 |
0.6608 | 20.99 | 850 | 0.3586 | 0.1821 | 0.0529 |
0.6608 | 22.0 | 891 | 0.3541 | 0.1729 | 0.0526 |
0.5138 | 22.99 | 931 | 0.3346 | 0.1664 | 0.0499 |
0.5138 | 24.0 | 972 | 0.3342 | 0.1586 | 0.0484 |
0.4831 | 24.99 | 1012 | 0.3314 | 0.1564 | 0.0484 |
0.4831 | 26.0 | 1053 | 0.3205 | 0.1521 | 0.0464 |
0.4831 | 26.99 | 1093 | 0.3114 | 0.1464 | 0.0460 |
0.4154 | 28.0 | 1134 | 0.3047 | 0.14 | 0.0441 |
0.4154 | 28.99 | 1174 | 0.3021 | 0.1364 | 0.0434 |
0.3806 | 30.0 | 1215 | 0.2956 | 0.135 | 0.0430 |
0.3806 | 30.99 | 1255 | 0.2861 | 0.1321 | 0.0422 |
0.3806 | 32.0 | 1296 | 0.2739 | 0.1271 | 0.0402 |
0.3351 | 32.99 | 1336 | 0.2791 | 0.1293 | 0.0413 |
0.3351 | 34.0 | 1377 | 0.2691 | 0.1271 | 0.0399 |
0.3066 | 34.99 | 1417 | 0.2650 | 0.1286 | 0.0392 |
0.3066 | 36.0 | 1458 | 0.2645 | 0.1293 | 0.0398 |
0.3066 | 36.99 | 1498 | 0.2602 | 0.1271 | 0.0399 |
0.2792 | 38.0 | 1539 | 0.2601 | 0.1293 | 0.0399 |
0.2792 | 38.99 | 1579 | 0.2562 | 0.1229 | 0.0388 |
0.3037 | 40.0 | 1620 | 0.2501 | 0.1264 | 0.0384 |
0.3037 | 40.99 | 1660 | 0.2466 | 0.1193 | 0.0373 |
0.2447 | 42.0 | 1701 | 0.2474 | 0.1186 | 0.0374 |
0.2447 | 42.99 | 1741 | 0.2422 | 0.1179 | 0.0373 |
0.2447 | 44.0 | 1782 | 0.2446 | 0.1193 | 0.0387 |
0.2362 | 44.99 | 1822 | 0.2408 | 0.115 | 0.0361 |
0.2362 | 46.0 | 1863 | 0.2439 | 0.1179 | 0.0374 |
0.2633 | 46.99 | 1903 | 0.2459 | 0.1221 | 0.0382 |
0.2633 | 48.0 | 1944 | 0.2478 | 0.1271 | 0.0392 |
0.2633 | 48.99 | 1984 | 0.2375 | 0.1164 | 0.0368 |
0.2204 | 50.0 | 2025 | 0.2379 | 0.1207 | 0.0378 |
0.2204 | 50.99 | 2065 | 0.2416 | 0.1214 | 0.0388 |
0.2186 | 52.0 | 2106 | 0.2383 | 0.1179 | 0.0368 |
0.2186 | 52.99 | 2146 | 0.2366 | 0.125 | 0.0375 |
0.2186 | 54.0 | 2187 | 0.2328 | 0.1186 | 0.0364 |
0.2192 | 54.99 | 2227 | 0.2316 | 0.1143 | 0.0358 |
0.2192 | 56.0 | 2268 | 0.2346 | 0.1171 | 0.0370 |
0.199 | 56.99 | 2308 | 0.2273 | 0.1171 | 0.0354 |
0.199 | 58.0 | 2349 | 0.2338 | 0.1143 | 0.0361 |
0.199 | 58.99 | 2389 | 0.2323 | 0.1121 | 0.0343 |
0.206 | 60.0 | 2430 | 0.2346 | 0.1143 | 0.0353 |
0.206 | 60.99 | 2470 | 0.2328 | 0.1157 | 0.0360 |
0.1828 | 62.0 | 2511 | 0.2315 | 0.1121 | 0.0344 |
0.1828 | 62.99 | 2551 | 0.2347 | 0.1079 | 0.0346 |
0.1828 | 64.0 | 2592 | 0.2355 | 0.1093 | 0.0344 |
0.1866 | 64.99 | 2632 | 0.2363 | 0.1107 | 0.0346 |
0.1866 | 66.0 | 2673 | 0.2362 | 0.1086 | 0.0347 |
0.1743 | 66.99 | 2713 | 0.2350 | 0.1114 | 0.0347 |
0.1743 | 68.0 | 2754 | 0.2304 | 0.1079 | 0.0342 |
0.1743 | 68.99 | 2794 | 0.2344 | 0.1086 | 0.0346 |
0.1858 | 70.0 | 2835 | 0.2351 | 0.1064 | 0.0340 |
0.1858 | 70.99 | 2875 | 0.2368 | 0.1071 | 0.0344 |
0.1765 | 72.0 | 2916 | 0.2316 | 0.1114 | 0.0349 |
0.1765 | 72.99 | 2956 | 0.2330 | 0.1071 | 0.0337 |
0.1765 | 74.0 | 2997 | 0.2334 | 0.1043 | 0.0340 |
0.1702 | 74.99 | 3037 | 0.2357 | 0.1071 | 0.0350 |
0.1702 | 76.0 | 3078 | 0.2344 | 0.1079 | 0.0349 |
0.1745 | 76.99 | 3118 | 0.2350 | 0.1093 | 0.0350 |
Framework versions
- Transformers 4.28.0
- Pytorch 2.0.1+cu117
- Datasets 2.14.4
- Tokenizers 0.13.3