wav2vec2-augmented-vivos
This model is a fine-tuned version of facebook/wav2vec2-base on the vivos dataset. It achieves the following results on the evaluation set:
- Loss: 0.4403
- Wer: 0.2447
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: 0.0002
- train_batch_size: 32
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 128
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.3
- num_epochs: 20
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 6.6691 | 2.0 | 146 | 4.0324 | 1.0 |
| 3.4795 | 4.0 | 292 | 3.6294 | 1.0 |
| 3.4178 | 6.0 | 438 | 3.4910 | 1.0 |
| 1.8415 | 8.0 | 584 | 0.7926 | 0.5287 |
| 0.5336 | 10.0 | 730 | 0.5809 | 0.3677 |
| 0.3349 | 12.0 | 876 | 0.4679 | 0.2853 |
| 0.2424 | 14.0 | 1022 | 0.4440 | 0.2680 |
| 0.2193 | 16.0 | 1168 | 0.4420 | 0.2536 |
| 0.1627 | 18.0 | 1314 | 0.4373 | 0.2455 |
| 0.1532 | 20.0 | 1460 | 0.4403 | 0.2447 |
Framework versions
- Transformers 4.44.0
- Pytorch 2.4.0
- Datasets 2.21.0
- Tokenizers 0.19.1