nocturneFlow/whisper-small-common-augmented-kk

Whisper Small KK - Kazakh Augmented

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: 0.6225
• Wer: 46.5572

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: 100
• training_steps: 1000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.1921	1.5152	100	0.5052	47.7890
0.0705	3.0303	200	0.5226	49.1156
0.0077	4.5455	300	0.5783	47.1573
0.0035	6.0606	400	0.5758	46.0834
0.0008	7.5758	500	0.6011	46.4940
0.0006	9.0909	600	0.6126	46.3677
0.0003	10.6061	700	0.6159	44.4409
0.0002	12.1212	800	0.6194	44.3146
0.0002	13.6364	900	0.6216	46.5256
0.0002	15.1515	1000	0.6225	46.5572

Framework versions

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