LIA-AvignonUniversity

/

IWSLT2022-tamasheq-only

like 1

Hello. What do you mean by a vocab.json?

I think it is used by the Wav2Vec2Processor for tokenization. For example, this model has one https://huggingface.co/facebook/wav2vec2-base-960h

This shouldn't be necessary for a purely pre-trained model (no ASR). I find it a bit odd they included it.

So it was not fine-tuned on labeled data. Thank you for your prompt answer.