WhisperTiny15hCommonVoice30hAugment

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

• Loss: 0.8565
• Wer: 0.4172

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: 5000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.98	0.2976	500	1.0271	0.4861
0.8044	0.5952	1000	0.9400	0.4785
0.7403	0.8929	1500	0.9092	0.4431
0.575	1.1905	2000	0.8876	0.4370
0.5335	1.4881	2500	0.8789	0.4309
0.5165	1.7857	3000	0.8660	0.4287
0.4262	2.0833	3500	0.8623	0.4259
0.4143	2.3810	4000	0.8622	0.4222
0.4144	2.6786	4500	0.8545	0.4201
0.4047	2.9762	5000	0.8565	0.4172

Framework versions

• Transformers 4.48.3
• Pytorch 2.6.0
• Datasets 3.3.0
• Tokenizers 0.21.0