aisha-org/Whisper-Uzbek

Whisper Medium UZB - AISHA

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

• Loss: 0.2859
• Wer: 31.7790

Model description

More information needed

Intended uses & limitations

More information needed

Founder: Rifat Mamayusupov

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
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.5187	0.5392	1000	0.4935	44.1403
0.3423	1.0785	2000	0.4008	37.6948
0.3018	1.6177	3000	0.3739	36.3575
0.2401	2.1569	4000	0.2821	31.7791

Framework versions

• Transformers 4.41.2
• Pytorch 2.3.1+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1