deepdml/whisper-medium-af-fleurs-norm

Whisper Medium af

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

• Loss: 0.6584
• Wer: 24.1212
• Cer: 9.2133

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: 16
• eval_batch_size: 16
• 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: 600

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.322	0.1667	100	0.6194	25.8355	9.9346
0.0922	1.1183	200	0.6106	25.8528	10.0431
0.0363	2.07	300	0.6271	24.5714	10.3715
0.019	3.0217	400	0.6469	24.8831	10.5211
0.011	3.1883	500	0.6518	27.1861	11.8553
0.0056	4.14	600	0.6584	24.1212	9.2133

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