MohammadKhosravi/whisper-large-v3-Persian

Whisper large v3 - Mohammad Khosravi

This model is a fine-tuned version of openai/whisper-large-v3 on the Common Voice 11.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.2444
• Wer: 26.8658

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: 4
• 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: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.1817	0.4296	1000	0.3223	35.6097
0.1224	0.8591	2000	0.2781	31.9537
0.0703	1.2887	3000	0.2761	31.3946
0.057	1.7182	4000	0.2458	27.7885
0.0263	2.1478	5000	0.2444	26.8658

Framework versions

• Transformers 4.48.1
• Pytorch 2.5.1+cu121
• Datasets 3.2.0
• Tokenizers 0.21.0