hamna11/whisper-tiny-en-asian-accent

Whisper tiny En - Hamna Ramzan

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

• Loss: 0.3981
• Wer: 16.4164

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: 64
• 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
0.4438	0.25	1000	0.5724	23.7751
0.3441	1.1425	2000	0.4521	18.5095
0.2427	2.035	3000	0.4335	18.0589
0.2993	2.285	4000	0.3981	16.4164

Framework versions

• Transformers 4.52.0.dev0
• Pytorch 2.6.0+cu124
• Datasets 3.5.2.dev0
• Tokenizers 0.21.1