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

Whisper Large-V3 Basque

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

• Loss: 0.2533
• Wer: 7.8873

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
0.0099	11.1112	5000	0.1972	8.7187
0.0042	22.2225	10000	0.2275	8.6173
0.0045	33.3337	15000	0.2295	8.6103
0.0027	44.4449	20000	0.2422	8.0303
0.0016	55.5562	25000	0.2452	8.1326
0.0022	66.6674	30000	0.2578	8.6745
0.0005	77.7786	35000	0.2533	7.8873
0.001	88.8899	40000	0.2613	8.4465
0.0014	100.0	45000	0.2644	8.4604
0.0011	111.1112	50000	0.2592	8.2245
0.0003	122.2225	55000	0.2679	8.0745
0.0001	133.3337	60000	0.2671	7.9219

Framework versions

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