IbratDO/whisper-small-linkdata-tashkent-dialect

Whisper Small UZ - Link Data

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

• Loss: 0.5056
• Wer: 41.7576

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: 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
0.4746	1.2210	1000	0.5591	48.2297
0.3081	2.4420	2000	0.4815	42.7295
0.2342	3.6630	3000	0.4749	41.7905
0.1822	4.8840	4000	0.4863	41.4266
0.1224	6.1050	5000	0.5056	41.7576

Framework versions

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