zuazo/whisper-large-v3-turbo-eu-cv21.0

Whisper Large-V3-Turbo Basque

This model is a fine-tuned version of openai/whisper-large-v3-turbo on the common_voice_21_0_eu dataset. It achieves the following results on the evaluation set:

• Loss: 0.3890
• Wer: 7.9445

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: 3.75e-05
• train_batch_size: 32
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• 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: 100000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0191	11.1112	5000	0.2372	10.4892
0.0104	22.2225	10000	0.2676	10.2733
0.0048	33.3337	15000	0.2892	10.2039
0.0061	44.4449	20000	0.2959	10.0548
0.0052	55.5562	25000	0.3025	9.8909
0.0037	66.6674	30000	0.3136	10.5681
0.0026	77.7786	35000	0.3198	9.9664
0.0029	88.8899	40000	0.3295	10.6158
0.0014	100.0	45000	0.3219	9.8233
0.0007	111.1112	50000	0.3314	9.4045
0.0013	122.2225	55000	0.3390	9.9508
0.0004	133.3337	60000	0.3317	9.5042
0.0009	144.4449	65000	0.3369	9.2051
0.0003	155.5562	70000	0.3441	9.4540
0.0001	166.6674	75000	0.3372	8.9450
0.0	177.7786	80000	0.3462	8.9242
0.0	188.8899	85000	0.3559	8.7829
0.0	200.0	90000	0.3691	8.3572
0.0	211.1112	95000	0.3827	8.0503
0.0	222.2225	100000	0.3890	7.9445

Framework versions

• Transformers 4.52.3
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.1