Whisper Tiny es

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

Loss: 0.3749
Wer Raw: 21.6741
Cer Raw: 8.4585
Wer: 21.6747
Cer: 8.4585

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: 128
eval_batch_size: 128
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
0.3891	0.05	1000	0.6939	37.6291	14.8103	37.4810	14.7830
0.6597	0.1	2000	0.6883	34.8302	14.4460	34.8016	14.4405
0.1388	0.15	3000	0.6530	34.3358	13.1348	34.2976	13.1269
0.1217	0.2	4000	0.6682	34.5982	13.2295	34.5646	13.2233
0.1101	0.25	5000	0.6721	35.2325	13.7696	35.1969	13.7628
0.3564	0.3	6000	0.5374	28.6885	11.2075	28.6529	11.2010
0.3803	0.35	7000	0.3890	20.9394	7.6348	20.9369	7.6342
0.3373	0.4	8000	0.3540	19.9264	7.3914	19.9258	7.3913
0.2828	1.0134	9000	0.3343	18.6452	6.8455	18.6452	6.8455
0.2743	1.0634	10000	0.3218	17.8476	6.5983	17.8476	6.5983
0.2747	1.1134	11000	0.3123	17.2210	6.3677	17.2210	6.3677
0.116	1.1634	12000	0.3392	19.6398	7.3330	19.6398	7.3330
0.1081	1.2134	13000	0.3717	22.2626	8.5878	22.2626	8.5878
0.0918	1.2634	14000	0.3865	22.8498	8.6366	22.8498	8.6366
0.0908	1.3134	15000	0.3957	23.3710	8.8869	23.3710	8.8869
0.493	1.3634	16000	0.4018	22.0878	8.2129	22.0878	8.2129
0.4284	1.4134	17000	0.3943	22.6592	8.8754	22.6592	8.8754
0.3042	2.0268	18000	0.3967	22.3014	8.5581	22.3014	8.5581
0.2942	2.0768	19000	0.3946	22.5136	8.8319	22.5136	8.8319
0.1001	2.1268	20000	0.3749	21.6741	8.4585	21.6747	8.4585

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