wav2vec2-burak-v2.4-large
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice dataset. It achieves the following results on the evaluation set:
- Loss: 0.6419
- Wer: 0.4358
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: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- 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 |
|---|---|---|---|---|
| 4.8354 | 3.51 | 400 | 3.0971 | 1.0 |
| 1.4155 | 7.02 | 800 | 0.6770 | 0.6790 |
| 0.4583 | 10.52 | 1200 | 0.6025 | 0.5591 |
| 0.3014 | 14.03 | 1600 | 0.6563 | 0.5150 |
| 0.2231 | 17.54 | 2000 | 0.6393 | 0.4892 |
| 0.1738 | 21.05 | 2400 | 0.6653 | 0.4677 |
| 0.1451 | 24.56 | 2800 | 0.6167 | 0.4428 |
| 0.1277 | 28.07 | 3200 | 0.6419 | 0.4358 |
Framework versions
- Transformers 4.21.3
- Pytorch 1.12.1+cu113
- Datasets 2.4.0
- Tokenizers 0.12.1