whisper-small-dar
This model is a fine-tuned version of openai/whisper-small on the mozilla-foundation/common_voice_17_0 ar dataset. It achieves the following results on the evaluation set:
- Loss: 0.1813
- Wer: 0.3367
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
- distributed_type: multi-GPU
- num_devices: 2
- total_train_batch_size: 32
- total_eval_batch_size: 32
- optimizer: Use 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: 100
- training_steps: 1000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.303 | 0.5935 | 200 | 0.2434 | 0.4226 |
| 0.2 | 1.1869 | 400 | 0.2035 | 0.3914 |
| 0.1633 | 1.7804 | 600 | 0.1876 | 0.3469 |
| 0.106 | 2.3739 | 800 | 0.1850 | 0.3488 |
| 0.1005 | 2.9674 | 1000 | 0.1813 | 0.3367 |
Framework versions
- Transformers 4.48.0.dev0
- Pytorch 2.4.1+cu121
- Datasets 3.2.0
- Tokenizers 0.21.0