Where does this model come from?
Hello, from the original repo (https://github.com/facebookresearch/encodec), it seems like only 24kHz and 48kHz models are provided. May I know where does this 32kHz model come from?
Is this from the MusicGen paper? I see that the MusicGen checkpoints' config.json also points to this 32kHz model. It would be good if we have a model card for this model.
cc @sanchit-gandhi :)
Hey @gaunernst - that's correct. This 32kHz checkpoint is an artefact released as part of the MusicGen project. I'll add a README to this checkpoint to make that clearer, since this is missing currently.
Does the updated README help clarify this? https://huggingface.co/facebook/encodec_32khz
Feel free to open a PR to update any bits you'd like to see changed/updated!
Awesome! Thank you for the quick update!