deepdml/whisper-base-es-mix-norm

Whisper Base es

This model is a fine-tuned version of openai/whisper-base on the Common Voice 17.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.2606
• Wer Raw: 13.5692
• Cer Raw: 4.7682
• Wer: 13.5692
• Cer: 4.7682

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: 64
• seed: 42
• optimizer: Use 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_ratio: 0.04
• training_steps: 20000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Raw	Cer Raw	Wer	Cer
0.6462	0.05	1000	0.3777	19.3398	6.7897	19.2820	6.7778
0.2957	0.1	2000	0.3367	16.8778	5.8318	16.8689	5.8301
0.3301	0.15	3000	0.3260	16.4634	5.7942	16.4558	5.7930
0.2835	0.2	4000	0.3136	15.7885	5.5159	15.7840	5.5151
0.1578	0.25	5000	0.3031	15.2432	5.3172	15.2406	5.3168
0.1701	0.3	6000	0.2976	15.2400	5.3622	15.2394	5.3621
0.1496	0.35	7000	0.2908	14.9508	5.3064	14.9502	5.3063
0.162	0.4	8000	0.2873	14.5759	5.2136	14.5752	5.2135
0.1493	0.45	9000	0.2811	14.1876	5.0806	14.1869	5.0805
0.1897	0.5	10000	0.2799	14.1310	5.0065	14.1310	5.0065
0.1593	0.55	11000	0.2774	14.1272	4.9508	14.1272	4.9508
0.1656	0.6	12000	0.2704	14.0719	5.0714	14.0719	5.0714
0.2724	0.65	13000	0.2724	13.9181	4.9589	13.9181	4.9587
0.1613	0.7	14000	0.2675	13.8240	4.8865	13.8240	4.8865
0.1206	0.75	15000	0.2694	13.4770	4.7620	13.4770	4.7620
0.1168	0.8	16000	0.2640	13.4542	4.7793	13.4542	4.7793
0.1432	0.85	17000	0.2640	13.5076	4.7499	13.5076	4.7499
0.1729	0.9	18000	0.2632	13.5037	4.7636	13.5037	4.7636
0.1083	0.95	19000	0.2609	13.5813	4.8009	13.5813	4.8009
0.1209	1.0	20000	0.2606	13.5692	4.7682	13.5692	4.7682

Framework versions

• Transformers 4.48.0.dev0
• Pytorch 2.5.1+cu121
• Datasets 3.6.0
• Tokenizers 0.21.0

Citation

Please cite the model using the following BibTeX entry:

@misc{deepdml/whisper-base-es-mix-norm,
      title={Fine-tuned Whisper base ASR model for speech recognition in Spanish},
      author={Jimenez, David},
      howpublished={\url{https://huggingface.co/deepdml/whisper-base-es-mix-norm}},
      year={2026}
    }