Mahmoud-Nasser/whisper-small-ar

Whisper Base Ar - GPTeam

This model is a fine-tuned version of openai/whisper-base on the quranic_audio_dataset dataset. It achieves the following results on the evaluation set:

• Loss: 0.0527
• Wer: 29.2050

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: 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
• training_steps: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0771	2.9240	1000	0.0722	34.2806
0.0183	5.8480	2000	0.0553	30.8476
0.0062	8.7719	3000	0.0527	30.7654
0.0023	11.6959	4000	0.0527	29.2050

Framework versions

• Transformers 4.49.0
• Pytorch 2.5.1+cu124
• Datasets 3.3.2
• Tokenizers 0.21.0