whisper-medium-dora-afrispeech-all
This model is a fine-tuned version of openai/whisper-medium on the afrispeech-200 dataset. It achieves the following results on the evaluation set:
- Loss: 0.8189
- Model Preparation Time: 0.0449
- Wer: 0.125
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: 0.0001
- train_batch_size: 16
- eval_batch_size: 16
- 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: 3589
- num_epochs: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Model Preparation Time | Wer |
|---|---|---|---|---|---|
| 0.5406 | 1.0 | 3590 | 0.6143 | 0.0449 | 0.139 |
| 0.4742 | 2.0 | 7180 | 0.5470 | 0.0449 | 0.126 |
| 0.3209 | 3.0 | 10770 | 0.5525 | 0.0449 | 0.125 |
| 0.2135 | 4.0 | 14360 | 0.5756 | 0.0449 | 0.126 |
| 0.1865 | 5.0 | 17950 | 0.6595 | 0.0449 | 0.126 |
| 0.1048 | 6.0 | 21540 | 0.8189 | 0.0449 | 0.125 |
Framework versions
- PEFT 0.18.1
- Transformers 5.3.0
- Pytorch 2.6.0+cu124
- Datasets 2.19.0
- Tokenizers 0.22.2