anton-l's picture
anton-l HF staff
Upload README.md
2da02d1
metadata
language:
  - nl
license: apache-2.0
tags:
  - automatic-speech-recognition
  - mozilla-foundation/common_voice_8_0
  - generated_from_trainer
  - nl
  - robust-speech-event
  - model_for_talk
  - hf-asr-leaderboard
datasets:
  - mozilla-foundation/common_voice_8_0
model-index:
  - name: wav2vec2-large-xls-r-300m-cv8-nl
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: Common Voice 8
          type: mozilla-foundation/common_voice_8_0
          args: nl
        metrics:
          - name: Test WER
            type: wer
            value: 14.53
          - name: Test CER
            type: cer
            value: 4.7
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: Robust Speech Event - Dev Data
          type: speech-recognition-community-v2/dev_data
          args: nl
        metrics:
          - name: Test WER
            type: wer
            value: 33.7
          - name: Test CER
            type: cer
            value: 15.64
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: Robust Speech Event - Test Data
          type: speech-recognition-community-v2/eval_data
          args: nl
        metrics:
          - name: Test WER
            type: wer
            value: 35.19

wav2vec2-large-xls-r-300m-cv8-nl

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice dataset. In addition a 6gram KenLM model was trained and used. The KenLM model was based on train+validation Common Voice 8 It achieves results depicted on the rigth side on the model card (testset CV8)

Model description

Dutch wav2vec2-xls-r-300m model using Common Voice 8 dataset

Intended uses & limitations

More information needed

Training and evaluation data

The model was trained on Dutch common voice 8 with 75 epochs. The train set consisted of the common voice 8 train set and evaluation set was the common voice 8 validation set. The WER reported is on the common voice 8 test set which was not part of training nor validation (eval)

Training procedure

Training hyperparameters

Framework versions

  • Transformers 4.16.0.dev0
  • Pytorch 1.10.1+cu102
  • Datasets 1.18.1
  • Tokenizers 0.11.0