nocturneFlow/whisper-medium-common-kk

Whisper Medium Kk - Kazakh

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: 0.5977
• Wer: 37.9027

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 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: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0078	15.1515	1000	0.5202	41.3771
0.0001	30.3030	2000	0.5584	39.6715
0.0	45.4545	3000	0.5884	38.0606
0.0	60.6061	4000	0.5977	37.9027

Framework versions

• Transformers 4.51.3
• Pytorch 2.6.0+cu118
• Datasets 3.6.0
• Tokenizers 0.21.0