IbratDO/whisper-large-linkdata-tashkent-dialect

Whisper large-v3 UZ - Link Data

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

• Loss: 0.4566
• Wer: 34.2610

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: 4
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED 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
0.306	1.2212	1000	0.3767	37.0396
0.1833	2.4424	2000	0.3431	34.0597
0.1262	3.6636	3000	0.3548	33.7957
0.0755	4.8848	4000	0.3905	34.0269
0.0284	6.1051	5000	0.4566	34.2610

Framework versions

• Transformers 4.57.0
• Pytorch 2.8.0+cu126
• Datasets 4.1.1
• Tokenizers 0.22.1