wav2vec2-bert-cv16-mas-ex-cv16
This model is a fine-tuned version of facebook/w2v-bert-2.0 on the common_voice_16_0 dataset. It achieves the following results on the evaluation set:
- Loss: 0.7349
- Wer: 0.6612
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: 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: 700
- num_epochs: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 1.3593 | 1.21 | 700 | 0.6050 | 0.5216 |
| 0.5443 | 2.43 | 1400 | 0.5665 | 0.4557 |
| 0.9415 | 3.64 | 2100 | 0.6099 | 0.5665 |
| 1.0953 | 4.85 | 2800 | 0.7349 | 0.6612 |
| 1.176 | 6.07 | 3500 | 0.7349 | 0.6612 |
| 1.1783 | 7.28 | 4200 | 0.7349 | 0.6612 |
| 1.1771 | 8.49 | 4900 | 0.7349 | 0.6612 |
| 1.1775 | 9.71 | 5600 | 0.7349 | 0.6612 |
Framework versions
- Transformers 4.39.1
- Pytorch 2.2.0
- Datasets 2.15.0
- Tokenizers 0.15.2