zuazo/whisper-small-eu-train

Whisper Small Basque

This model is a fine-tuned version of openai/whisper-small on the mozilla-foundation/common_voice_13_0 eu dataset. It achieves the following results on the evaluation set:

• Loss: 0.3485
• Wer: 19.3516

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: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.337	5.85	1000	0.3838	26.8815
0.1384	11.7	2000	0.3431	22.1544
0.0732	17.54	3000	0.3376	20.0447
0.0432	23.39	4000	0.3458	19.5349
0.0378	29.24	5000	0.3485	19.3516

Framework versions

• Transformers 4.37.2
• Pytorch 2.2.0+cu121
• Datasets 2.16.1
• Tokenizers 0.15.1