README.md

metadata

language:
  - fr
license: apache-2.0
tags:
  - whisper
  - generated_from_trainer
datasets:
  - mozilla-foundation/common_voice_15_0
  - BrunoHays/multilingual-tedx-fr
  - PolyAI/minds14
  - facebook/multilingual_librispeech
  - facebook/voxpopuli
  - google/fleurs
metrics:
  - wer
model-index:
  - name: Whisper tiny French
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset1:
          name: mozilla-foundation/common_voice_15_0 fr
          type: mozilla-foundation/common_voice_15_0
          config: fr
          split: test
          args: fr
        metrics:
          - name: Wer
            type: wer
            value: 40
        dataset2:
          name: facebook/multilingual_librispeech fr
          type: facebook/multilingual_librispeech
          config: fr
          split: test
          args: fr
          wer: 26.1
        dataset3:
          name: facebook/voxpopuli fr
          type: facebook/voxpopuli
          config: fr
          split: test
          args: fr
          wer: 29.4
        dataset4:
          name: google/fleurs fr
          type: google/fleurs
          config: fr
          split: test
          args: fr
          wer: 33.7

Whisper tiny fr - JaepaX

This model is a fine-tuned version of openai/whisper-tiny on the fr datasets.

WER Result

It achieves the following results on the evaluation sets

• Mulit-Libri : "26.1",
• common : "40.0"
• voxpopuli : "29.4"
• fleurs : "33.7"