deepdml/whisper-large-v3-turbo-ar-mix-norm

Whisper Turbo ar

This model is a fine-tuned version of deepdml/whisper-large-v3-turbo on the Common Voice 17.0 dataset. It achieves the following results on the evaluation set:

• Loss: 2.3375
• Wer: 13.8006

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: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 5000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.4123	0.2	1000	2.2390	19.6509
0.2413	0.4	2000	2.3998	17.6051
0.1324	0.6	3000	2.3918	16.3601
0.0796	0.8	4000	2.3270	14.7044
0.0311	1.0	5000	2.3375	13.8006

Framework versions

• Transformers 4.42.0.dev0
• Pytorch 2.3.0+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1

Citation

@misc{deepdml/whisper-large-v3-turbo-ar-mix-norm,
      title={Fine-tuned Whisper turbo ASR model for speech recognition in Arabic},
      author={Jimenez, David},
      howpublished={\url{https://huggingface.co/deepdml/whisper-large-v3-turbo-ar-mix-norm}},
      year={2025}
    }