gencgeray/whisper-tiny-uk

Whisper tiny uk - Herai Hench KI-11

This model is a fine-tuned version of openai/whisper-tiny on the Common Voice 11.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.7245
• Wer: 58.3932
• Cer: 18.1853

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: 6e-06
• 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: 6000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.7374	0.8065	1000	0.8454	64.6981	22.8643
0.6193	1.6129	2000	0.7735	61.6387	20.3717
0.5334	2.4194	3000	0.7444	60.3618	18.8322
0.4709	3.2258	4000	0.7318	59.5903	19.9146
0.4616	4.0323	5000	0.7242	58.3134	18.0517
0.4209	4.8387	6000	0.7245	58.3932	18.1853

Framework versions

• Transformers 4.52.4
• Pytorch 2.5.1+cu124
• Datasets 3.6.0
• Tokenizers 0.21.0