KalamTech/whisper-small-ar-cv-11

Whisper Small AR - Mohammed Bakheet

This model is a fine-tuned version of openai/whisper-small on the Common Voice 11.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.2601
• Wer: 20.4562

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: 8
• seed: 42
• gradient_accumulation_steps: 16
• 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
• training_steps: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.5279	0.4158	500	0.3311	27.6591
0.2513	0.8316	1000	0.2866	24.5504
0.1673	1.2478	1500	0.2735	22.8928
0.1324	1.6635	2000	0.2645	21.8153
0.1138	2.0797	2500	0.2613	21.3816
0.064	2.4955	3000	0.2651	21.0006
0.0615	2.9113	3500	0.2601	20.4562

Framework versions

• Transformers 4.46.2
• Pytorch 2.5.1+cu124
• Datasets 3.1.0
• Tokenizers 0.20.3