WhisperTiny15hCommonVoice45hAugmentV2

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: 0.8412
• Wer: 0.4180

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
0.9528	0.2233	500	0.9953	0.4696
0.7794	0.4466	1000	0.9161	0.4818
0.6888	0.6699	1500	0.8845	0.4304
0.5905	0.8933	2000	0.8664	0.4301
0.4504	1.1166	2500	0.8553	0.4267
0.4521	1.3399	3000	0.8520	0.4231
0.4061	1.5632	3500	0.8458	0.4194
0.39	1.7865	4000	0.8475	0.4179
0.3882	2.0098	4500	0.8440	0.4168
0.3237	2.2331	5000	0.8412	0.4180

Framework versions

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