torgo-sentences-TestedSpeakerM01
This model is a fine-tuned version of yongjian/wav2vec2-large-a on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.4934
- Wer: 0.7531
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: 4
- 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: 1000
- num_epochs: 30
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 16.1147 | 3.36 | 500 | 0.7890 | 0.9519 |
| 0.6317 | 6.71 | 1000 | 0.5087 | 0.8789 |
| 0.5158 | 10.07 | 1500 | 0.5307 | 0.9519 |
| 0.4783 | 13.42 | 2000 | 0.4320 | 0.9208 |
| 0.5256 | 16.78 | 2500 | 0.3832 | 0.8292 |
| 0.2892 | 20.13 | 3000 | 0.4282 | 0.7531 |
| 0.2733 | 23.49 | 3500 | 0.4369 | 0.7391 |
| 0.2289 | 26.85 | 4000 | 0.4782 | 0.7469 |
Framework versions
- Transformers 4.23.1
- Pytorch 1.12.1+cu113
- Datasets 1.18.3
- Tokenizers 0.13.2