<!-- 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-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: 0.1182
- Wer: 0.0856
- Cer: 0.0243
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 |
---|---|---|---|---|---|
13.9933 | 0.98 | 24 | 5.9190 | 1.0 | 1.0 |
13.9933 | 2.0 | 49 | 3.3838 | 1.0 | 1.0 |
13.9933 | 2.98 | 73 | 3.1066 | 1.0 | 1.0 |
13.9933 | 4.0 | 98 | 2.9911 | 1.0 | 1.0 |
5.3123 | 4.98 | 122 | 2.9350 | 1.0 | 1.0 |
5.3123 | 6.0 | 147 | 2.9081 | 1.0 | 1.0 |
5.3123 | 6.98 | 171 | 2.8878 | 1.0 | 1.0 |
5.3123 | 8.0 | 196 | 2.8781 | 1.0 | 1.0 |
2.9259 | 8.98 | 220 | 2.8696 | 1.0 | 1.0 |
2.9259 | 10.0 | 245 | 2.8702 | 1.0 | 1.0 |
2.9259 | 10.98 | 269 | 2.8564 | 1.0 | 1.0 |
2.9259 | 12.0 | 294 | 2.8394 | 1.0 | 1.0 |
2.8604 | 12.98 | 318 | 2.7822 | 1.0 | 1.0 |
2.8604 | 14.0 | 343 | 2.5910 | 1.0 | 1.0 |
2.8604 | 14.98 | 367 | 2.2933 | 0.9994 | 0.9998 |
2.8604 | 16.0 | 392 | 1.6601 | 1.0 | 0.5076 |
2.5379 | 16.98 | 416 | 1.0895 | 0.9572 | 0.2661 |
2.5379 | 18.0 | 441 | 0.7374 | 0.4191 | 0.1002 |
2.5379 | 18.98 | 465 | 0.5591 | 0.2985 | 0.0719 |
2.5379 | 20.0 | 490 | 0.4534 | 0.2396 | 0.0577 |
1.145 | 20.98 | 514 | 0.4009 | 0.2277 | 0.0548 |
1.145 | 22.0 | 539 | 0.3531 | 0.2021 | 0.0515 |
1.145 | 22.98 | 563 | 0.3164 | 0.1932 | 0.0483 |
1.145 | 24.0 | 588 | 0.2922 | 0.1813 | 0.0464 |
0.6153 | 24.98 | 612 | 0.2722 | 0.1724 | 0.0451 |
0.6153 | 26.0 | 637 | 0.2508 | 0.1688 | 0.0427 |
0.6153 | 26.98 | 661 | 0.2394 | 0.1605 | 0.0409 |
0.6153 | 28.0 | 686 | 0.2263 | 0.1540 | 0.0399 |
0.4548 | 28.98 | 710 | 0.2185 | 0.1480 | 0.0384 |
0.4548 | 30.0 | 735 | 0.2072 | 0.1409 | 0.0368 |
0.4548 | 30.98 | 759 | 0.1953 | 0.1302 | 0.0334 |
0.4548 | 32.0 | 784 | 0.1872 | 0.1159 | 0.0311 |
0.4105 | 32.98 | 808 | 0.1838 | 0.1183 | 0.0317 |
0.4105 | 34.0 | 833 | 0.1844 | 0.1112 | 0.0310 |
0.4105 | 34.98 | 857 | 0.1742 | 0.1100 | 0.0301 |
0.4105 | 36.0 | 882 | 0.1683 | 0.1106 | 0.0299 |
0.3478 | 36.98 | 906 | 0.1652 | 0.1088 | 0.0298 |
0.3478 | 38.0 | 931 | 0.1611 | 0.1112 | 0.0302 |
0.3478 | 38.98 | 955 | 0.1578 | 0.1058 | 0.0290 |
0.3478 | 40.0 | 980 | 0.1566 | 0.1094 | 0.0290 |
0.3056 | 40.98 | 1004 | 0.1555 | 0.1106 | 0.0297 |
0.3056 | 42.0 | 1029 | 0.1554 | 0.1058 | 0.0294 |
0.3056 | 42.98 | 1053 | 0.1517 | 0.1046 | 0.0284 |
0.3056 | 44.0 | 1078 | 0.1497 | 0.1046 | 0.0282 |
0.2838 | 44.98 | 1102 | 0.1475 | 0.1023 | 0.0275 |
0.2838 | 46.0 | 1127 | 0.1422 | 0.0945 | 0.0264 |
0.2838 | 46.98 | 1151 | 0.1411 | 0.0945 | 0.0263 |
0.2838 | 48.0 | 1176 | 0.1410 | 0.0945 | 0.0265 |
0.2722 | 48.98 | 1200 | 0.1393 | 0.0927 | 0.0266 |
0.2722 | 50.0 | 1225 | 0.1389 | 0.0910 | 0.0263 |
0.2722 | 50.98 | 1249 | 0.1370 | 0.0951 | 0.0266 |
0.2722 | 52.0 | 1274 | 0.1370 | 0.0927 | 0.0266 |
0.2722 | 52.98 | 1298 | 0.1355 | 0.0933 | 0.0265 |
0.2349 | 54.0 | 1323 | 0.1340 | 0.0910 | 0.0257 |
0.2349 | 54.98 | 1347 | 0.1351 | 0.0922 | 0.0264 |
0.2349 | 56.0 | 1372 | 0.1322 | 0.0927 | 0.0258 |
0.2349 | 56.98 | 1396 | 0.1328 | 0.0904 | 0.0255 |
0.2357 | 58.0 | 1421 | 0.1317 | 0.0910 | 0.0258 |
0.2357 | 58.98 | 1445 | 0.1313 | 0.0898 | 0.0253 |
0.2357 | 60.0 | 1470 | 0.1301 | 0.0904 | 0.0259 |
0.2357 | 60.98 | 1494 | 0.1296 | 0.0892 | 0.0254 |
0.227 | 62.0 | 1519 | 0.1291 | 0.0892 | 0.0258 |
0.227 | 62.98 | 1543 | 0.1293 | 0.0927 | 0.0264 |
0.227 | 64.0 | 1568 | 0.1266 | 0.0910 | 0.0259 |
0.227 | 64.98 | 1592 | 0.1257 | 0.0904 | 0.0253 |
0.2162 | 66.0 | 1617 | 0.1261 | 0.0910 | 0.0260 |
0.2162 | 66.98 | 1641 | 0.1257 | 0.0910 | 0.0259 |
0.2162 | 68.0 | 1666 | 0.1266 | 0.0939 | 0.0263 |
0.2162 | 68.98 | 1690 | 0.1254 | 0.0916 | 0.0261 |
0.2165 | 70.0 | 1715 | 0.1249 | 0.0880 | 0.0258 |
0.2165 | 70.98 | 1739 | 0.1237 | 0.0910 | 0.0253 |
0.2165 | 72.0 | 1764 | 0.1228 | 0.0874 | 0.0249 |
0.2165 | 72.98 | 1788 | 0.1228 | 0.0874 | 0.0249 |
0.192 | 74.0 | 1813 | 0.1225 | 0.0862 | 0.0249 |
0.192 | 74.98 | 1837 | 0.1222 | 0.0880 | 0.0250 |
0.192 | 76.0 | 1862 | 0.1207 | 0.0868 | 0.0245 |
0.192 | 76.98 | 1886 | 0.1210 | 0.0856 | 0.0242 |
0.2059 | 78.0 | 1911 | 0.1204 | 0.0844 | 0.0240 |
0.2059 | 78.98 | 1935 | 0.1199 | 0.0862 | 0.0243 |
0.2059 | 80.0 | 1960 | 0.1191 | 0.0850 | 0.0242 |
0.2059 | 80.98 | 1984 | 0.1186 | 0.0862 | 0.0243 |
0.1814 | 82.0 | 2009 | 0.1189 | 0.0856 | 0.0243 |
0.1814 | 82.98 | 2033 | 0.1186 | 0.0862 | 0.0247 |
0.1814 | 84.0 | 2058 | 0.1187 | 0.0862 | 0.0243 |
0.1814 | 84.98 | 2082 | 0.1182 | 0.0856 | 0.0243 |
0.1894 | 86.0 | 2107 | 0.1185 | 0.0856 | 0.0244 |
0.1894 | 86.98 | 2131 | 0.1184 | 0.0832 | 0.0240 |
0.1894 | 88.0 | 2156 | 0.1183 | 0.0850 | 0.0242 |
0.1894 | 88.98 | 2180 | 0.1184 | 0.0826 | 0.0237 |
0.1784 | 90.0 | 2205 | 0.1187 | 0.0844 | 0.0242 |
0.1784 | 90.98 | 2229 | 0.1190 | 0.0844 | 0.0241 |
0.1784 | 92.0 | 2254 | 0.1193 | 0.0856 | 0.0244 |
0.1784 | 92.98 | 2278 | 0.1193 | 0.0850 | 0.0243 |
0.1941 | 94.0 | 2303 | 0.1190 | 0.0850 | 0.0243 |
0.1941 | 94.98 | 2327 | 0.1188 | 0.0844 | 0.0243 |
0.1941 | 96.0 | 2352 | 0.1187 | 0.0820 | 0.0236 |
0.1941 | 96.98 | 2376 | 0.1186 | 0.0820 | 0.0236 |
0.1886 | 97.96 | 2400 | 0.1187 | 0.0820 | 0.0236 |
Framework versions
- Transformers 4.28.0
- Pytorch 2.0.1+cu117
- Datasets 2.14.0
- Tokenizers 0.13.3