wav2vec2_base_2
This model is a fine-tuned version of facebook/wav2vec2-base on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.1226
- Wer: 0.1541
- Cer: 0.0436
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: 0.0001
- train_batch_size: 16
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 7
Training results
Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
---|---|---|---|---|---|
5.0885 | 0.2 | 700 | 2.9120 | 1.0 | 0.9992 |
1.7483 | 0.41 | 1400 | 0.5196 | 0.4141 | 0.1443 |
0.7658 | 0.61 | 2100 | 0.3747 | 0.3313 | 0.1119 |
0.6498 | 0.81 | 2800 | 0.3081 | 0.2879 | 0.0936 |
0.5693 | 1.02 | 3500 | 0.2744 | 0.2618 | 0.0841 |
0.4756 | 1.22 | 4200 | 0.2429 | 0.2366 | 0.0769 |
0.4488 | 1.43 | 4900 | 0.2355 | 0.2192 | 0.0705 |
0.4614 | 1.63 | 5600 | 0.2337 | 0.2185 | 0.0700 |
0.4064 | 1.83 | 6300 | 0.2015 | 0.2044 | 0.0642 |
0.392 | 2.04 | 7000 | 0.1966 | 0.2014 | 0.0629 |
0.3606 | 2.24 | 7700 | 0.1957 | 0.1956 | 0.0605 |
0.355 | 2.44 | 8400 | 0.1895 | 0.1891 | 0.0583 |
0.3294 | 2.65 | 9100 | 0.1767 | 0.1891 | 0.0577 |
0.3224 | 2.85 | 9800 | 0.1678 | 0.1874 | 0.0571 |
0.3066 | 3.05 | 10500 | 0.1671 | 0.1855 | 0.0568 |
0.2859 | 3.26 | 11200 | 0.1740 | 0.1825 | 0.0552 |
0.2758 | 3.46 | 11900 | 0.1551 | 0.1808 | 0.0542 |
0.2696 | 3.66 | 12600 | 0.1615 | 0.1773 | 0.0535 |
0.2663 | 3.87 | 13300 | 0.1431 | 0.1728 | 0.0515 |
0.2527 | 4.07 | 14000 | 0.1381 | 0.1697 | 0.0500 |
0.2375 | 4.28 | 14700 | 0.1436 | 0.1680 | 0.0493 |
0.2529 | 4.48 | 15400 | 0.1470 | 0.1645 | 0.0483 |
0.2374 | 4.68 | 16100 | 0.1329 | 0.1657 | 0.0480 |
0.2362 | 4.89 | 16800 | 0.1293 | 0.1639 | 0.0476 |
0.2205 | 5.09 | 17500 | 0.1466 | 0.1630 | 0.0471 |
0.2177 | 5.29 | 18200 | 0.1348 | 0.1619 | 0.0469 |
0.1983 | 5.5 | 18900 | 0.1262 | 0.1598 | 0.0455 |
0.1936 | 5.7 | 19600 | 0.1248 | 0.1584 | 0.0452 |
0.2055 | 5.9 | 20300 | 0.1252 | 0.1583 | 0.0450 |
0.1824 | 6.11 | 21000 | 0.1247 | 0.1564 | 0.0446 |
0.1931 | 6.31 | 21700 | 0.1217 | 0.1552 | 0.0440 |
0.1857 | 6.52 | 22400 | 0.1213 | 0.1562 | 0.0442 |
0.1929 | 6.72 | 23100 | 0.1214 | 0.1544 | 0.0437 |
0.1648 | 6.92 | 23800 | 0.1226 | 0.1541 | 0.0436 |
Framework versions
- Transformers 4.39.3
- Pytorch 2.1.2
- Datasets 2.18.0
- Tokenizers 0.15.2
- Downloads last month
- 2