<!-- 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-07
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: 0.1190
- Wer: 0.0838
- Cer: 0.0235
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 |
---|---|---|---|---|---|
26.9545 | 1.0 | 67 | 3.3218 | 1.0 | 1.0 |
6.7922 | 2.0 | 134 | 2.9660 | 1.0 | 1.0 |
2.9975 | 3.0 | 201 | 2.9543 | 1.0 | 1.0 |
2.9975 | 4.0 | 268 | 2.9085 | 1.0 | 1.0 |
2.9164 | 5.0 | 335 | 2.8689 | 1.0 | 1.0 |
2.8317 | 6.0 | 402 | 2.4003 | 0.9993 | 0.8377 |
2.8317 | 7.0 | 469 | 0.8135 | 0.4399 | 0.1061 |
1.665 | 8.0 | 536 | 0.4696 | 0.2409 | 0.0607 |
0.8096 | 9.0 | 603 | 0.3550 | 0.2042 | 0.0527 |
0.8096 | 10.0 | 670 | 0.2970 | 0.1890 | 0.0482 |
0.6279 | 11.0 | 737 | 0.2574 | 0.1616 | 0.0428 |
0.4796 | 12.0 | 804 | 0.2268 | 0.1568 | 0.0409 |
0.4796 | 13.0 | 871 | 0.2154 | 0.1488 | 0.0400 |
0.4053 | 14.0 | 938 | 0.1954 | 0.1340 | 0.0358 |
0.3627 | 15.0 | 1005 | 0.1850 | 0.1274 | 0.0345 |
0.3627 | 16.0 | 1072 | 0.1694 | 0.1177 | 0.0312 |
0.3498 | 17.0 | 1139 | 0.1631 | 0.1094 | 0.0299 |
0.2956 | 18.0 | 1206 | 0.1555 | 0.1056 | 0.0287 |
0.2956 | 19.0 | 1273 | 0.1522 | 0.1035 | 0.0290 |
0.2792 | 20.0 | 1340 | 0.1527 | 0.1076 | 0.0296 |
0.2767 | 21.0 | 1407 | 0.1453 | 0.1014 | 0.0276 |
0.2767 | 22.0 | 1474 | 0.1433 | 0.1031 | 0.0287 |
0.2511 | 23.0 | 1541 | 0.1396 | 0.0969 | 0.0273 |
0.2427 | 24.0 | 1608 | 0.1364 | 0.0973 | 0.0270 |
0.2427 | 25.0 | 1675 | 0.1346 | 0.0962 | 0.0261 |
0.2286 | 26.0 | 1742 | 0.1382 | 0.0931 | 0.0265 |
0.2202 | 27.0 | 1809 | 0.1331 | 0.0959 | 0.0261 |
0.2202 | 28.0 | 1876 | 0.1299 | 0.0917 | 0.0257 |
0.2363 | 29.0 | 1943 | 0.1260 | 0.0907 | 0.0251 |
0.1919 | 30.0 | 2010 | 0.1262 | 0.0893 | 0.0248 |
0.1919 | 31.0 | 2077 | 0.1260 | 0.0907 | 0.0254 |
0.1957 | 32.0 | 2144 | 0.1241 | 0.0907 | 0.0248 |
0.1969 | 33.0 | 2211 | 0.1254 | 0.0924 | 0.0248 |
0.1969 | 34.0 | 2278 | 0.1256 | 0.0886 | 0.0248 |
0.2182 | 35.0 | 2345 | 0.1228 | 0.0872 | 0.0248 |
0.1871 | 36.0 | 2412 | 0.1211 | 0.0872 | 0.0235 |
0.1871 | 37.0 | 2479 | 0.1231 | 0.0893 | 0.0245 |
0.1656 | 38.0 | 2546 | 0.1219 | 0.0886 | 0.0244 |
0.1678 | 39.0 | 2613 | 0.1190 | 0.0838 | 0.0235 |
0.1678 | 40.0 | 2680 | 0.1273 | 0.0907 | 0.0254 |
0.191 | 41.0 | 2747 | 0.1207 | 0.0869 | 0.0234 |
0.1772 | 42.0 | 2814 | 0.1231 | 0.0831 | 0.0235 |
0.1772 | 43.0 | 2881 | 0.1217 | 0.0876 | 0.0243 |
0.1786 | 44.0 | 2948 | 0.1244 | 0.0879 | 0.0244 |
0.1655 | 45.0 | 3015 | 0.1224 | 0.0876 | 0.0241 |
0.1655 | 46.0 | 3082 | 0.1254 | 0.0876 | 0.0244 |
0.1581 | 47.0 | 3149 | 0.1236 | 0.0907 | 0.0249 |
0.1567 | 48.0 | 3216 | 0.1208 | 0.0879 | 0.0241 |
0.1567 | 49.0 | 3283 | 0.1257 | 0.0914 | 0.0250 |
0.1489 | 50.0 | 3350 | 0.1221 | 0.0897 | 0.0252 |
0.1644 | 51.0 | 3417 | 0.1217 | 0.0869 | 0.0243 |
0.1644 | 52.0 | 3484 | 0.1245 | 0.0910 | 0.0253 |
0.1468 | 53.0 | 3551 | 0.1195 | 0.0862 | 0.0237 |
0.1461 | 54.0 | 3618 | 0.1198 | 0.0869 | 0.0235 |
0.1461 | 55.0 | 3685 | 0.1192 | 0.0879 | 0.0243 |
0.1435 | 56.0 | 3752 | 0.1242 | 0.0862 | 0.0242 |
0.1436 | 57.0 | 3819 | 0.1211 | 0.0893 | 0.0246 |
0.1436 | 58.0 | 3886 | 0.1230 | 0.0855 | 0.0243 |
0.1462 | 59.0 | 3953 | 0.1222 | 0.0897 | 0.0243 |
Framework versions
- Transformers 4.28.0
- Pytorch 2.0.0+cu117
- Datasets 2.12.0
- Tokenizers 0.13.3