<!-- 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-clean-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: 0.0926
- Wer: 0.0794
- Cer: 0.0206
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 |
---|---|---|---|---|---|
15.8761 | 1.0 | 23 | 5.4500 | 1.0 | 1.0 |
15.8761 | 2.0 | 46 | 3.3423 | 1.0 | 1.0 |
15.8761 | 3.0 | 69 | 3.0481 | 1.0 | 1.0 |
15.8761 | 4.0 | 92 | 2.9605 | 1.0 | 1.0 |
4.9341 | 5.0 | 115 | 2.9110 | 1.0 | 1.0 |
4.9341 | 6.0 | 138 | 2.8886 | 1.0 | 1.0 |
4.9341 | 7.0 | 161 | 2.8759 | 1.0 | 1.0 |
4.9341 | 8.0 | 184 | 2.8792 | 1.0 | 1.0 |
2.911 | 9.0 | 207 | 2.8496 | 1.0 | 1.0 |
2.911 | 10.0 | 230 | 2.8546 | 1.0 | 1.0 |
2.911 | 11.0 | 253 | 2.8693 | 1.0 | 1.0 |
2.911 | 12.0 | 276 | 2.8325 | 1.0 | 1.0 |
2.911 | 13.0 | 299 | 2.8073 | 1.0 | 1.0 |
2.8673 | 14.0 | 322 | 2.7405 | 1.0 | 1.0 |
2.8673 | 15.0 | 345 | 2.5617 | 1.0 | 1.0 |
2.8673 | 16.0 | 368 | 2.1579 | 1.0 | 0.8058 |
2.8673 | 17.0 | 391 | 1.4867 | 0.9124 | 0.3745 |
2.5054 | 18.0 | 414 | 0.9883 | 0.4345 | 0.1073 |
2.5054 | 19.0 | 437 | 0.6926 | 0.3282 | 0.0757 |
2.5054 | 20.0 | 460 | 0.5380 | 0.2349 | 0.0554 |
2.5054 | 21.0 | 483 | 0.4459 | 0.2021 | 0.0510 |
1.134 | 22.0 | 506 | 0.3795 | 0.1939 | 0.0478 |
1.134 | 23.0 | 529 | 0.3437 | 0.1915 | 0.0475 |
1.134 | 24.0 | 552 | 0.3055 | 0.1833 | 0.0438 |
1.134 | 25.0 | 575 | 0.2794 | 0.1710 | 0.0413 |
1.134 | 26.0 | 598 | 0.2607 | 0.1727 | 0.0410 |
0.6806 | 27.0 | 621 | 0.2422 | 0.1645 | 0.0387 |
0.6806 | 28.0 | 644 | 0.2296 | 0.1628 | 0.0396 |
0.6806 | 29.0 | 667 | 0.2151 | 0.1522 | 0.0376 |
0.6806 | 30.0 | 690 | 0.2024 | 0.1498 | 0.0364 |
0.47 | 31.0 | 713 | 0.1909 | 0.1408 | 0.0347 |
0.47 | 32.0 | 736 | 0.1838 | 0.1318 | 0.0334 |
0.47 | 33.0 | 759 | 0.1742 | 0.1293 | 0.0317 |
0.47 | 34.0 | 782 | 0.1675 | 0.1195 | 0.0295 |
0.403 | 35.0 | 805 | 0.1605 | 0.1146 | 0.0282 |
0.403 | 36.0 | 828 | 0.1568 | 0.1170 | 0.0294 |
0.403 | 37.0 | 851 | 0.1525 | 0.1129 | 0.0279 |
0.403 | 38.0 | 874 | 0.1477 | 0.1137 | 0.0278 |
0.403 | 39.0 | 897 | 0.1442 | 0.1080 | 0.0275 |
0.3661 | 40.0 | 920 | 0.1411 | 0.1072 | 0.0265 |
0.3661 | 41.0 | 943 | 0.1382 | 0.0949 | 0.0248 |
0.3661 | 42.0 | 966 | 0.1349 | 0.0990 | 0.0243 |
0.3661 | 43.0 | 989 | 0.1315 | 0.0933 | 0.0240 |
0.3094 | 44.0 | 1012 | 0.1274 | 0.0933 | 0.0233 |
0.3094 | 45.0 | 1035 | 0.1252 | 0.0941 | 0.0240 |
0.3094 | 46.0 | 1058 | 0.1245 | 0.0917 | 0.0230 |
0.3094 | 47.0 | 1081 | 0.1214 | 0.0900 | 0.0233 |
0.2733 | 48.0 | 1104 | 0.1188 | 0.0892 | 0.0235 |
0.2733 | 49.0 | 1127 | 0.1163 | 0.0892 | 0.0236 |
0.2733 | 50.0 | 1150 | 0.1144 | 0.0917 | 0.0238 |
0.2733 | 51.0 | 1173 | 0.1159 | 0.0908 | 0.0232 |
0.2733 | 52.0 | 1196 | 0.1153 | 0.0908 | 0.0230 |
0.2607 | 53.0 | 1219 | 0.1129 | 0.0884 | 0.0232 |
0.2607 | 54.0 | 1242 | 0.1109 | 0.0876 | 0.0230 |
0.2607 | 55.0 | 1265 | 0.1103 | 0.0867 | 0.0229 |
0.2607 | 56.0 | 1288 | 0.1122 | 0.0908 | 0.0242 |
0.242 | 57.0 | 1311 | 0.1101 | 0.0867 | 0.0230 |
0.242 | 58.0 | 1334 | 0.1085 | 0.0884 | 0.0235 |
0.242 | 59.0 | 1357 | 0.1095 | 0.0917 | 0.0238 |
0.242 | 60.0 | 1380 | 0.1081 | 0.0876 | 0.0223 |
0.2301 | 61.0 | 1403 | 0.1075 | 0.0859 | 0.0229 |
0.2301 | 62.0 | 1426 | 0.1064 | 0.0908 | 0.0236 |
0.2301 | 63.0 | 1449 | 0.1047 | 0.0876 | 0.0232 |
0.2301 | 64.0 | 1472 | 0.1025 | 0.0884 | 0.0230 |
0.2301 | 65.0 | 1495 | 0.1000 | 0.0827 | 0.0222 |
0.2264 | 66.0 | 1518 | 0.0978 | 0.0859 | 0.0225 |
0.2264 | 67.0 | 1541 | 0.0999 | 0.0835 | 0.0220 |
0.2264 | 68.0 | 1564 | 0.1006 | 0.0843 | 0.0222 |
0.2264 | 69.0 | 1587 | 0.0994 | 0.0827 | 0.0225 |
0.2327 | 70.0 | 1610 | 0.0983 | 0.0827 | 0.0223 |
0.2327 | 71.0 | 1633 | 0.0967 | 0.0827 | 0.0220 |
0.2327 | 72.0 | 1656 | 0.0968 | 0.0810 | 0.0217 |
0.2327 | 73.0 | 1679 | 0.0997 | 0.0802 | 0.0217 |
0.2051 | 74.0 | 1702 | 0.0998 | 0.0835 | 0.0228 |
0.2051 | 75.0 | 1725 | 0.0979 | 0.0843 | 0.0217 |
0.2051 | 76.0 | 1748 | 0.0977 | 0.0810 | 0.0219 |
0.2051 | 77.0 | 1771 | 0.0969 | 0.0777 | 0.0212 |
0.2051 | 78.0 | 1794 | 0.0985 | 0.0810 | 0.0217 |
0.1829 | 79.0 | 1817 | 0.0962 | 0.0777 | 0.0213 |
0.1829 | 80.0 | 1840 | 0.0963 | 0.0786 | 0.0212 |
0.1829 | 81.0 | 1863 | 0.0956 | 0.0802 | 0.0212 |
0.1829 | 82.0 | 1886 | 0.0951 | 0.0810 | 0.0210 |
0.1851 | 83.0 | 1909 | 0.0943 | 0.0786 | 0.0209 |
0.1851 | 84.0 | 1932 | 0.0947 | 0.0794 | 0.0210 |
0.1851 | 85.0 | 1955 | 0.0955 | 0.0810 | 0.0213 |
0.1851 | 86.0 | 1978 | 0.0936 | 0.0818 | 0.0213 |
0.2014 | 87.0 | 2001 | 0.0941 | 0.0802 | 0.0210 |
0.2014 | 88.0 | 2024 | 0.0943 | 0.0794 | 0.0207 |
0.2014 | 89.0 | 2047 | 0.0943 | 0.0810 | 0.0213 |
0.2014 | 90.0 | 2070 | 0.0928 | 0.0794 | 0.0209 |
0.2014 | 91.0 | 2093 | 0.0927 | 0.0786 | 0.0206 |
0.1904 | 92.0 | 2116 | 0.0926 | 0.0794 | 0.0206 |
0.1904 | 93.0 | 2139 | 0.0928 | 0.0802 | 0.0212 |
0.1904 | 94.0 | 2162 | 0.0929 | 0.0810 | 0.0215 |
0.1904 | 95.0 | 2185 | 0.0929 | 0.0802 | 0.0210 |
0.1864 | 96.0 | 2208 | 0.0929 | 0.0794 | 0.0207 |
0.1864 | 97.0 | 2231 | 0.0932 | 0.0802 | 0.0210 |
0.1864 | 98.0 | 2254 | 0.0929 | 0.0794 | 0.0207 |
0.1864 | 99.0 | 2277 | 0.0930 | 0.0802 | 0.0209 |
0.189 | 100.0 | 2300 | 0.0931 | 0.0802 | 0.0209 |
Framework versions
- Transformers 4.28.0
- Pytorch 2.0.1+cu117
- Datasets 2.14.4
- Tokenizers 0.13.3