Whisper Tiny Ar - AzeemX
This model is a fine-tuned version of openai/whisper-tiny on the Tashkeel dataset. It achieves the following results on the evaluation set:
- Loss: 0.2374
- Wer: 40.0
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: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- training_steps: 5000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.2521 | 1.6 | 1000 | 0.3009 | 53.0811 |
| 0.146 | 3.2 | 2000 | 0.2476 | 42.5946 |
| 0.1238 | 4.8 | 3000 | 0.2334 | 40.1081 |
| 0.0916 | 6.4 | 4000 | 0.2372 | 39.5315 |
| 0.0866 | 8.0 | 5000 | 0.2374 | 40.0 |
Framework versions
- Transformers 4.45.1
- Pytorch 2.4.1+cu121
- Datasets 3.0.1
- Tokenizers 0.20.0