xezpeleta/whisper-base-eu

Whisper Base Basque

This model is a fine-tuned version of openai/whisper-base on the asierhv/composite_corpus_eu_v2.1 dataset. It achieves the following results on the evaluation set:

• Loss: 0.2452
• Wer: 13.8170

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: 2e-05
• train_batch_size: 32
• eval_batch_size: 16
• 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: 8000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.4951	0.125	1000	0.4901	27.0543
0.2607	0.25	2000	0.3708	19.8654
0.1887	0.375	3000	0.3454	18.3977
0.2607	0.5	4000	0.3218	16.8085
0.106	0.625	5000	0.3289	15.7801
0.1376	0.75	6000	0.3052	15.0276
0.1733	0.875	7000	0.3004	13.9338
0.1228	1.0	8000	0.2452	13.8170

Framework versions

• Transformers 4.49.0.dev0
• Pytorch 2.6.0+cu124
• Datasets 3.2.1.dev0
• Tokenizers 0.21.0