README.md

metadata

language:
  - ar
tags:
  - whisper-event
  - generated_from_trainer
datasets:
  - mozilla-foundation/common_voice_11_0
metrics:
  - wer
model-index:
  - name: Whisper Medium AR - gerryc
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: FLEURS
          type: google/fleurs
          config: ar_eg
          split: test
          args: ar
        metrics:
          - name: Wer
            type: wer
            value: 52.8

Whisper Medium Ar - gerryc

Model was trained on CommonVoice Train. Tensorboard eval is on 256 samples of CommonVoice Dev.

Not normalized or lowercased. Evaluated and saved every 2500 steps. This model was overtrained and is overfitted. About 2000 to 4000 steps is around the best spot to stop training.