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: 0.0954
- Wer: 0.2010
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 |
|---|---|---|---|---|
| 2.3079 | 1.0 | 313 | 0.0917 | 0.2065 |
| 2.1753 | 2.0 | 626 | 0.0927 | 0.1955 |
| 1.7907 | 3.0 | 939 | 0.0925 | 0.1986 |
| 1.504 | 4.0 | 1252 | 0.0927 | 0.2017 |
| 1.3552 | 5.0 | 1565 | 0.0926 | 0.1989 |
| 1.2117 | 6.0 | 1878 | 0.0920 | 0.1933 |
| 1.0949 | 7.0 | 2191 | 0.0910 | 0.1981 |
| 0.9967 | 8.0 | 2504 | 0.0915 | 0.2006 |
| 0.9329 | 9.0 | 2817 | 0.0920 | 0.1981 |
| 0.8878 | 10.0 | 3130 | 0.0907 | 0.1997 |
| 0.8541 | 11.0 | 3443 | 0.0903 | 0.2022 |
| 0.7744 | 12.0 | 3756 | 0.0906 | 0.2007 |
| 0.7645 | 13.0 | 4069 | 0.0900 | 0.1985 |
| 0.7268 | 14.0 | 4382 | 0.0900 | 0.1995 |
| 0.702 | 14.9536 | 4680 | 0.0900 | 0.1997 |
Framework versions
- Transformers 4.51.1
- Pytorch 2.5.1+cu124
- Datasets 3.5.0
- Tokenizers 0.21.0