wav2vec2-large-xls-r-300m-dysarthria-big-dataset
This model is a fine-tuned version of jmaczan/wav2vec2-large-xls-r-300m-dysarthria-big-dataset on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.0864
- Wer: 0.182
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: 0.0003
- train_batch_size: 16
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 30
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 1.419 | 3.2 | 200 | 0.7599 | 0.668 |
| 0.7759 | 6.4 | 400 | 0.4966 | 0.618 |
| 0.5808 | 9.6 | 600 | 0.3352 | 0.508 |
| 0.3652 | 12.8 | 800 | 0.2214 | 0.386 |
| 0.2347 | 16.0 | 1000 | 0.1566 | 0.246 |
| 0.1738 | 19.2 | 1200 | 0.1340 | 0.23 |
| 0.1076 | 22.4 | 1400 | 0.1244 | 0.242 |
| 0.077 | 25.6 | 1600 | 0.0948 | 0.184 |
| 0.0566 | 28.8 | 1800 | 0.0864 | 0.182 |
Framework versions
- Transformers 4.43.2
- Pytorch 2.2.1+cu121
- Datasets 2.20.0
- Tokenizers 0.19.1