Whisper-Sinhala_Audio_to_Text
This model is a fine-tuned version of openai/whisper-small on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.9038
- Wer: 50.0822
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: 1e-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: 500
- num_epochs: 50
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.0665 | 4.76 | 1000 | 0.5398 | 57.8125 |
| 0.0096 | 9.52 | 2000 | 0.6716 | 56.2089 |
| 0.0037 | 14.29 | 3000 | 0.7457 | 52.7549 |
| 0.0005 | 19.05 | 4000 | 0.8000 | 51.1513 |
| 0.002 | 23.81 | 5000 | 0.8057 | 51.6859 |
| 0.0005 | 28.57 | 6000 | 0.8150 | 50.3289 |
| 0.0005 | 33.33 | 7000 | 0.8445 | 51.0280 |
| 0.0 | 38.1 | 8000 | 0.8773 | 50.1234 |
| 0.0 | 42.86 | 9000 | 0.8944 | 50.1234 |
| 0.0 | 47.62 | 10000 | 0.9038 | 50.0822 |
Framework versions
- Transformers 4.38.2
- Pytorch 2.2.1+cu121
- Datasets 2.18.0
- Tokenizers 0.15.2