w2v-bert-2.0-sr
This model is a fine-tuned version of facebook/w2v-bert-2.0 on the common_voice_16_1 dataset. It achieves the following results on the evaluation set:
- Loss: 0.1469
- Wer: 0.0534
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: 5e-05
- train_batch_size: 4
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 8
- 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: 20
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 2.1994 | 1.89 | 300 | 0.1350 | 0.1078 |
| 0.2331 | 3.77 | 600 | 0.2306 | 0.1341 |
| 0.1879 | 5.66 | 900 | 0.1354 | 0.0766 |
| 0.1579 | 7.54 | 1200 | 0.1646 | 0.0958 |
| 0.1293 | 9.43 | 1500 | 0.1207 | 0.0713 |
| 0.1182 | 11.31 | 1800 | 0.1376 | 0.0737 |
| 0.1061 | 13.2 | 2100 | 0.1244 | 0.0580 |
| 0.1011 | 15.08 | 2400 | 0.1390 | 0.0602 |
| 0.0933 | 16.97 | 2700 | 0.1313 | 0.0524 |
| 0.0948 | 18.85 | 3000 | 0.1469 | 0.0534 |
Framework versions
- Transformers 4.37.2
- Pytorch 2.1.2+cu121
- Datasets 2.16.1
- Tokenizers 0.15.1