Yorkinjon/whisper-small-uzbek-ynv3

Whisper Small uz - Yorkerdev

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

• Loss: 0.3481
• Wer: 33.6874

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: 2
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 16
• optimizer: Use 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: 8000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.6132	0.2640	1000	0.5668	49.3073
0.4946	0.5280	2000	0.4544	40.9201
0.4363	0.7919	3000	0.4098	37.8935
0.3124	1.0557	4000	0.3872	38.4938
0.2919	1.3197	5000	0.3674	35.4522
0.2711	1.5837	6000	0.3577	34.6721
0.2754	1.8476	7000	0.3491	33.6874
0.2024	2.1114	8000	0.3481	33.6874

Framework versions

• Transformers 4.49.0
• Pytorch 2.6.0+cu118
• Datasets 3.3.2
• Tokenizers 0.21.0