Point users to the official bosonai/higgs-audio-v2-tokenizer?
Hi @eustlb , thanks a lot for porting the Higgs Audio v2 tokenizer to transformers. It’s been really helpful for the community.
We’ve now updated the official repo bosonai/higgs-audio-v2-tokenizer so its model.safetensors / config.json match yours. It works natively with transformers>=5.3.0 and is the one we’ll maintain going forward (license, docs, updates).
Many projects (e.g. k2-fsa/OmniVoice and its forks) still load from this repo. Would you be open to one of these?
Adding a short notice at the top of this model card pointing to bosonai/higgs-audio-v2-tokenizer as the canonical repo, and/or
If HF supports it internally, setting up a redirect from this repo to ours.
Happy to do whatever is easiest for you. My email address is ke@boson.ai. Feel free to contact me if you have any other concern. Thanks again!
— Boson AI team
Hey, thanks for reaching out, I did not know this repo was getting this much usage!
What I did:
- added a banner here to clearly point to the official repo
- opened https://github.com/k2-fsa/OmniVoice/pull/297 and https://huggingface.co/k2-fsa/OmniVoice/discussions/40/files to point
omnivoiceto correct repo
Thank you so much for the prompt reply and help fixing all the issues!