wav2vec2-base-twi-demo
This model is a fine-tuned version of facebook/wav2vec2-base on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.0564
- Wer: 0.3246
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: 8
- eval_batch_size: 8
- seed: 42
- optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 1000
- num_epochs: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 11.0853 | 0.1781 | 500 | 3.1694 | 1.0 |
| 3.1605 | 0.3563 | 1000 | 3.1995 | 1.0 |
| 3.0661 | 0.5344 | 1500 | 2.8477 | 1.0 |
| 1.6835 | 0.7125 | 2000 | 0.7489 | 0.6972 |
| 0.7354 | 0.8906 | 2500 | 0.4461 | 0.5712 |
| 0.5237 | 1.0688 | 3000 | 0.3285 | 0.5233 |
| 0.4167 | 1.2469 | 3500 | 0.2644 | 0.4881 |
| 0.3423 | 1.4250 | 4000 | 0.2429 | 0.4500 |
| 0.3162 | 1.6031 | 4500 | 0.2356 | 0.4322 |
| 0.267 | 1.7813 | 5000 | 0.2025 | 0.4306 |
| 0.2795 | 1.9594 | 5500 | 0.1766 | 0.4198 |
| 0.2181 | 2.1375 | 6000 | 0.1932 | 0.3943 |
| 0.2166 | 2.3156 | 6500 | 0.1578 | 0.3817 |
| 0.2236 | 2.4938 | 7000 | 0.2271 | 0.3955 |
| 0.2142 | 2.6719 | 7500 | 0.1487 | 0.3796 |
| 0.2025 | 2.8500 | 8000 | 0.1558 | 0.3731 |
| 0.1914 | 3.0281 | 8500 | 0.1298 | 0.3728 |
| 0.1623 | 3.2063 | 9000 | 0.3313 | 0.4452 |
| 0.188 | 3.3844 | 9500 | 0.1381 | 0.3709 |
| 0.1783 | 3.5625 | 10000 | 0.1177 | 0.3601 |
| 0.1365 | 3.7406 | 10500 | 0.1307 | 0.3693 |
| 0.146 | 3.9188 | 11000 | 0.1203 | 0.3645 |
| 0.1393 | 4.0969 | 11500 | 0.1096 | 0.3572 |
| 0.139 | 4.2750 | 12000 | 0.1065 | 0.3531 |
| 0.1271 | 4.4532 | 12500 | 0.0969 | 0.3553 |
| 0.1522 | 4.6313 | 13000 | 0.1120 | 0.3679 |
| 0.1441 | 4.8094 | 13500 | 0.1129 | 0.3570 |
| 0.1113 | 4.9875 | 14000 | 0.1033 | 0.3525 |
| 0.1115 | 5.1657 | 14500 | 0.0960 | 0.3489 |
| 0.1171 | 5.3438 | 15000 | 0.0851 | 0.3479 |
| 0.1064 | 5.5219 | 15500 | 0.0855 | 0.3446 |
| 0.085 | 5.7000 | 16000 | 0.0913 | 0.3434 |
| 0.0817 | 5.8782 | 16500 | 0.0836 | 0.3456 |
| 0.0973 | 6.0563 | 17000 | 0.0693 | 0.3388 |
| 0.0854 | 6.2344 | 17500 | 0.0769 | 0.3396 |
| 0.0846 | 6.4125 | 18000 | 0.0846 | 0.3404 |
| 0.076 | 6.5907 | 18500 | 0.0895 | 0.3368 |
| 0.08 | 6.7688 | 19000 | 0.0813 | 0.3396 |
| 0.0735 | 6.9469 | 19500 | 0.0790 | 0.3369 |
| 0.0634 | 7.1250 | 20000 | 0.0795 | 0.3351 |
| 0.0714 | 7.3032 | 20500 | 0.0739 | 0.3361 |
| 0.0776 | 7.4813 | 21000 | 0.0801 | 0.3335 |
| 0.0695 | 7.6594 | 21500 | 0.0687 | 0.3353 |
| 0.0556 | 7.8375 | 22000 | 0.0689 | 0.3345 |
| 0.0567 | 8.0157 | 22500 | 0.0634 | 0.3310 |
| 0.0566 | 8.1938 | 23000 | 0.0631 | 0.3322 |
| 0.0453 | 8.3719 | 23500 | 0.0605 | 0.3279 |
| 0.0583 | 8.5501 | 24000 | 0.0599 | 0.3281 |
| 0.0475 | 8.7282 | 24500 | 0.0672 | 0.3299 |
| 0.0518 | 8.9063 | 25000 | 0.0621 | 0.3273 |
| 0.0402 | 9.0844 | 25500 | 0.0654 | 0.3270 |
| 0.0418 | 9.2626 | 26000 | 0.0605 | 0.3260 |
| 0.0451 | 9.4407 | 26500 | 0.0589 | 0.3279 |
| 0.0473 | 9.6188 | 27000 | 0.0600 | 0.3271 |
| 0.0456 | 9.7969 | 27500 | 0.0579 | 0.3250 |
| 0.036 | 9.9751 | 28000 | 0.0564 | 0.3246 |
Framework versions
- Transformers 4.46.3
- Pytorch 2.5.1+cu121
- Datasets 3.1.0
- Tokenizers 0.20.3