Whisper Small Amharic - Biniyam Daniel
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: 0.3777
- Wer: 51.4327
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: 8
- seed: 42
- optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- training_steps: 4000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.0866 | 4.1322 | 1000 | 0.2203 | 52.3233 |
| 0.0107 | 8.2645 | 2000 | 0.3035 | 52.2556 |
| 0.0008 | 12.3967 | 3000 | 0.3588 | 51.0552 |
| 0.0006 | 16.5289 | 4000 | 0.3777 | 51.4327 |
Framework versions
- Transformers 4.57.0
- Pytorch 2.6.0+cu124
- Datasets 3.6.0
- Tokenizers 0.22.1
- Downloads last month
- 12
Model tree for b1n1yam/shook
Base model
openai/whisper-small