whisper-small-arabic

This model is a fine-tuned version of openai/whisper-small on an badrex/arabic-speech-SADA22-MSA. It achieves the following results on the evaluation set:

• Loss: 0.6831
• Wer: 37.5697

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: 2
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 4
• 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: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.4906	1.0858	1000	0.6512	44.8384
0.2482	2.1716	2000	0.6218	40.3344
0.1214	3.2575	3000	0.6319	39.5318
0.0484	4.3433	4000	0.6658	38.0602
0.0256	5.4291	5000	0.6831	37.5697

Framework versions

• Transformers 4.52.3
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.1