Whisper Small bn - Shams
This model is a fine-tuned version of openai/whisper-small on the kaggle_comp dataset. It achieves the following results on the evaluation set:
- Loss: 0.4243
- Wer Ortho: 68.9499
- Wer: 43.7776
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: 16
- eval_batch_size: 16
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: constant_with_warmup
- lr_scheduler_warmup_steps: 5
- num_epochs: 5
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer Ortho | Wer |
|---|---|---|---|---|---|
| 0.365 | 1.5974 | 500 | 0.4108 | 75.3776 | 49.6947 |
| 0.2016 | 3.1949 | 1000 | 0.3938 | 71.2451 | 46.3694 |
| 0.14 | 4.7923 | 1500 | 0.4243 | 68.9499 | 43.7776 |
Framework versions
- Transformers 4.41.2
- Pytorch 2.3.0+cu118
- Datasets 2.20.0
- Tokenizers 0.19.1
- Downloads last month
- 5