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: 0.3069
Wer Raw: 17.3220
Cer Raw: 6.4775
Wer: 17.3220
Cer: 6.4775

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_steps: 0.04
training_steps: 26000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Raw	Cer Raw	Wer	Cer
0.3392	0.0385	1000	0.5449	28.8875	10.7007	28.7842	10.6805
0.2970	0.0769	2000	0.4834	25.9414	9.4732	25.9167	9.4688
0.2931	0.1154	3000	0.4402	24.4542	9.0891	24.4377	9.0862
0.3371	0.1538	4000	0.4287	23.8031	8.9298	23.7936	8.9282
0.4178	0.1923	5000	0.4111	23.1951	8.6842	23.1925	8.6838
0.2361	0.2308	6000	0.3823	20.7530	7.6529	20.7524	7.6528
0.3117	0.2692	7000	0.3746	21.4333	8.3745	21.4314	8.3741
0.3198	0.3077	8000	0.3682	20.9369	7.8742	20.9369	7.8742
0.2277	1.0035	9000	0.3481	19.8135	7.4420	19.8135	7.4420
0.2065	1.0420	10000	0.3423	19.0578	7.0356	19.0578	7.0356
0.1846	1.0804	11000	0.3353	18.7769	6.9491	18.7769	6.9491
0.1821	1.1189	12000	0.3343	18.3040	6.6649	18.3040	6.6649
0.2689	1.1573	13000	0.3397	19.2632	7.2140	19.2632	7.2140
0.2155	1.1958	14000	0.3329	18.8315	7.1981	18.8315	7.1981
0.1951	1.2343	15000	0.3257	18.3072	6.7254	18.3072	6.7254
0.3108	1.2727	16000	0.3247	18.5639	7.0022	18.5639	7.0022
0.3006	1.3112	17000	0.3245	17.9553	6.5605	17.9553	6.5605
0.1631	2.007	18000	0.3128	17.6130	6.6351	17.6130	6.6351
0.1956	2.0455	19000	0.3122	17.9471	6.7893	17.9471	6.7893
0.1764	2.0839	20000	0.3151	17.6967	6.5664	17.6967	6.5664
0.2162	2.1224	21000	0.3135	17.6808	6.6306	17.6808	6.6306
0.1736	2.1608	22000	0.3110	17.2941	6.3855	17.2941	6.3855
0.1488	2.1993	23000	0.3088	17.4982	6.4258	17.4982	6.4258
0.3267	2.2378	24000	0.3105	17.6624	6.6477	17.6624	6.6477
0.1671	2.2762	25000	0.3077	17.4634	6.5241	17.4634	6.5241
0.1411	2.3147	26000	0.3069	17.3220	6.4775	17.3220	6.4775

Framework versions

Transformers 5.14.1
Pytorch 2.6.0+cu124
Datasets 5.0.1
Tokenizers 0.22.2

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}
    }