metadata
library_name: transformers
language:
- ar
license: apache-2.0
base_model: openai/whisper-base
tags:
- generated_from_trainer
metrics:
- wer
model-index:
- name: Whisper base AR - BA
results: []
Whisper base AR - BA
This model is a fine-tuned version of openai/whisper-base on the quran-ayat-speech-to-text dataset. It achieves the following results on the evaluation set:
- Loss: 0.0875
- Wer: 0.1939
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: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 32
- 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
- num_epochs: 15
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 1.9262 | 1.0 | 469 | 0.0907 | 0.1917 |
| 1.8739 | 2.0 | 938 | 0.0907 | 0.1874 |
| 1.4395 | 3.0 | 1407 | 0.0898 | 0.1991 |
| 1.3835 | 4.0 | 1876 | 0.0880 | 0.2003 |
| 1.1837 | 5.0 | 2345 | 0.0884 | 0.2065 |
| 1.0963 | 6.0 | 2814 | 0.0870 | 0.2020 |
| 1.0446 | 7.0 | 3283 | 0.0877 | 0.1967 |
| 0.8945 | 8.0 | 3752 | 0.0867 | 0.1895 |
| 0.8261 | 9.0 | 4221 | 0.0858 | 0.1986 |
| 0.7863 | 10.0 | 4690 | 0.0849 | 0.1889 |
| 0.7641 | 11.0 | 5159 | 0.0854 | 0.1932 |
| 0.6898 | 12.0 | 5628 | 0.0853 | 0.1910 |
| 0.6968 | 13.0 | 6097 | 0.0845 | 0.1914 |
| 0.6905 | 14.0 | 6566 | 0.0846 | 0.1914 |
| 0.6316 | 14.9685 | 7020 | 0.0846 | 0.1858 |
Framework versions
- Transformers 4.51.3
- Pytorch 2.6.0+cu124
- Datasets 3.6.0
- Tokenizers 0.21.1