📄 Information about the usage of models and code with non-commercial LICENSE on transformers and diffusers
Hi Team,
I want to understand how HuggingFace handles code and models which are not licensed for commercial use in its repositories even though the repository itself is for commercial use. The transformers and diffusers libraries both are Apache License but both cater to non-commercial models as well.
I have read about model cards, and wanted to know if it is enough to just add the original license information in them along with mentioning the original license in the code headers or there are any extra steps being taken care of?
Also I believe the inference of these models would require code from the original authors itself, what if the original code is under a different license?
I am asking this question because I am building my first open source package mukh which is a Face Analysis library and the structure is inspired from HuggingFace where I provide a unified API for multiple models around Face Analysis tasks.
I have kept the repository under the Apache License but I have one model and code for that respective model which is under the GPL License.
Here are things I came across which can be done -
- Mention the original License in the code files and models being used, and use them in my repository with the current Apache License,
so it will be understood to a user that if they are using that particular model or it's code the repository license does not apply to it.
Convert my repository to GPL License
Remove the model from my repository and retain the Apache License
Is the first point valid? Or do I have to go ahead with 2 or 3?
Any suggestions are appreciated 🤗