wav2vec2-wtimit-finetune
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: 0.0383
- Wer: 0.0160
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: 64
- eval_batch_size: 16
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 1000
- num_epochs: 30
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 5.3743 | 2.82 | 500 | 2.9567 | 1.0 |
| 1.866 | 5.65 | 1000 | 0.2856 | 0.2580 |
| 0.2005 | 8.47 | 1500 | 0.0979 | 0.0669 |
| 0.08 | 11.3 | 2000 | 0.0617 | 0.0325 |
| 0.0497 | 14.12 | 2500 | 0.0578 | 0.0284 |
| 0.0348 | 16.95 | 3000 | 0.0557 | 0.0239 |
| 0.0269 | 19.77 | 3500 | 0.0447 | 0.0212 |
| 0.0198 | 22.6 | 4000 | 0.0437 | 0.0177 |
| 0.016 | 25.42 | 4500 | 0.0407 | 0.0164 |
| 0.014 | 28.25 | 5000 | 0.0383 | 0.0160 |
Framework versions
- Transformers 4.11.3
- Pytorch 1.9.1+cu111
- Datasets 1.13.3
- Tokenizers 0.10.3