Amazing work.
Thank you SO MUCH for this effort !
I was in the process of planning creating the same thing, and it was such an incredibly massive task...
You just saved me incredible amounts of work.
Thank you !!!
(FYI, I'm working on a series of coding-assistant LLMs created from code with specific licenses or copyright holders so people that contribute to open-source projects can use coding assistant without legal issues related to licenses/copyright... this will massively speed things up in terms of the first step of training, which has to be public domain as there is none of the necessary data in the target licenses)
If somebody happens to know of datasets of public domain (or sorted-by-license) code, I'd be very glad to learn about it too. One can dream...
@arthurwolf Perhaps BigCode's The Stack is what you're looking for? https://huggingface.co/datasets/bigcode/the-stack-dedup