README.md

metadata

library_name: transformers
language:
  - zu
license: apache-2.0
base_model: openai/whisper-small
tags:
  - generated_from_trainer
datasets:
  - Beijuka/nchlt_speech_corpus_ZULU_20
metrics:
  - wer
model-index:
  - name: Whisper Small Zu - Sibongile Manyate
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: NCHLT Zulu Speech corpus From Beijuka
          type: Beijuka/nchlt_speech_corpus_ZULU_20
        metrics:
          - name: Wer
            type: wer
            value: 45.627052669991485

Whisper Small Zu - Sibongile Manyate

This model is a fine-tuned version of openai/whisper-small on the NCHLT Zulu Speech corpus From Beijuka dataset. It achieves the following results on the evaluation set:

• Loss: 0.6037
• Wer Ortho: 46.8646
• Wer: 45.6271

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: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: constant_with_warmup
• lr_scheduler_warmup_steps: 50
• training_steps: 500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
0.5162	0.2532	500	0.6037	46.8646	45.6271

Framework versions

• Transformers 4.48.0
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0