We need to ensure that LLMs have a clear license. This is messy out there with a lot of new, weird, proprietary licenses.
There is something that would be super useful wrt. AI: find all these licenses!
The process would be:
When cloning all the repos @ Hugging Face, these should shallow clones, without Git LFS and without xet large files, and possibly partial clone to focus only on a few files: model cards and licenses, and related metadata files. See for ideas:
Running ScanCode license detection on that and review the results to find all the funky new AI-licenses and notices may be best done in ScanCode.io for efficiency, but the licenses are ultimately created in ScanCode toolkit
We need to ensure that LLMs have a clear license. This is messy out there with a lot of new, weird, proprietary licenses.
There is something that would be super useful wrt. AI: find all these licenses!
The process would be:
When cloning all the repos @ Hugging Face, these should shallow clones, without Git LFS and without xet large files, and possibly partial clone to focus only on a few files: model cards and licenses, and related metadata files. See for ideas:
Running ScanCode license detection on that and review the results to find all the funky new AI-licenses and notices may be best done in ScanCode.io for efficiency, but the licenses are ultimately created in ScanCode toolkit