WhisperTiny10hCommonVoice20hAugment

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: 0.9134
• Wer: 0.4361

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
0.9553	0.4492	500	1.0205	0.4894
0.7923	0.8985	1000	0.9367	0.4628
0.5788	1.3477	1500	0.9254	0.4439
0.5308	1.7969	2000	0.9195	0.4464
0.4003	2.2462	2500	0.9081	0.4373
0.3703	2.6954	3000	0.9046	0.4275
0.2858	3.1447	3500	0.9070	0.4381
0.2963	3.5939	4000	0.9081	0.4346
0.2686	4.0431	4500	0.9119	0.4346
0.2597	4.4924	5000	0.9134	0.4361

Framework versions

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