deepdml's picture
Update README.md
882e3d0
metadata
language:
  - af
license: apache-2.0
base_model: openai/whisper-base
tags:
  - generated_from_trainer
datasets:
  - google/fleurs
metrics:
  - wer
model-index:
  - name: Whisper Base af
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: Fleurs
          type: google/fleurs
          config: af_za
          split: test
          args: af_za
        metrics:
          - name: Wer
            type: wer
            value: 42.008658008658

Whisper Base af

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

  • Loss: 1.1171
  • Wer: 42.0087
  • Cer: 18.5731

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: 600

Training results

Training Loss Epoch Step Validation Loss Wer Cer
0.6138 3.0017 100 1.0982 48.7792 19.1919
0.2216 6.0033 200 1.0553 43.0303 17.0425
0.0984 9.005 300 1.0633 44.5195 19.8927
0.0527 12.0067 400 1.0950 45.1429 19.2446
0.0325 15.0083 500 1.1121 41.8182 16.4179
0.0291 18.01 600 1.1171 42.0087 18.5731

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