deepdml/whisper-tiny-ta-mix-norm

Whisper Tiny ta

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: 0.2583
• Wer: 50.9461
• Cer: 11.6470

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: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.04
• training_steps: 8000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.2315	0.125	1000	0.3384	62.1530	15.6705
0.1623	0.25	2000	0.2973	56.7273	13.4770
0.1716	0.375	3000	0.2856	55.2872	13.0946
0.1572	0.5	4000	0.2676	52.6516	12.2552
0.1475	0.625	5000	0.2650	52.0655	12.0396
0.1656	0.75	6000	0.2610	51.5322	11.9197
0.1048	0.875	7000	0.2561	50.7993	11.4955
0.1166	1.0	8000	0.2583	50.9461	11.6470

Framework versions

• Transformers 4.42.0.dev0
• Pytorch 2.3.0+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1

Citation

Please cite the model using the following BibTeX entry:

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