contrib : clarify code origin guidelines

[ddh0]

Overview

This PR updates the contributing guidelines to add a new section. At the time of posting, the section reads as follows:

## Code Origin Disclosure Policy

*Contributors* and *collaborators*, as defined above, must disclose the origin of their code, if they did not write it themselves, to avoid potential licensing conflicts or intellectual property right violations.

Contributions that can be shown to originate from one or more of the following sources, in whole or in part, will not be accepted:
- [ik_llama.cpp](https://github.com/ikawrakow/ik_llama.cpp)
- [Iwan Kawrakow](https://github.com/ikawrakow)

Note that maintainers reserve the right to reject code on the basis of its origin, even if not explicitly listed above.

Future readers are encouraged to view the diff of this PR, in case anything has changed. After this PR is merged, readers should instead refer to the current content of CONTRIBUTING.md, and should disregard this PR.

I also edited all Markdown headers except the first one from # to ##, such that the document now has a proper structure with a heading. I consider the new section to be the only substantial change.

Motivation

In #19726 and #21149 (and probably other instances), code was rejected the basis of its origin. In both cases, the contributors were not necessarily aware of the stance of llama.cpp with respect to what is allowed, and what is not.

I hope that this addition can clarify things for contributors, and reduce the amount of friction in the future. I welcome feedback from any maintainers, in particular @ggerganov and @JohannesGaessler. Thank you.

Requirements

• I have read and agree with the contributing guidelines
• AI usage disclosure: NO

[fairydreaming]

Contributions that can be shown to originate from one or more of the following sources, in whole or in part, will not be accepted:

• ik_llama.cpp

@ddh0 So let's say I'm implementing support for new model and it's merged in ik_llama.cpp first. Then I can't submit that PR to llama.cpp anymore?

[ddh0]

@ddh0 So let's say I'm implementing support for new model and it's merged in ik_llama.cpp first. Then I can't submit that PR to llama.cpp anymore?

I don't know - hence the PR and ping to core maintainers.

[0cc4m]

Contributions that can be shown to originate from one or more of the following sources, in whole or in part, will not be accepted:

• ik_llama.cpp

@ddh0 So let's say I'm implementing support for new model and it's merged in ik_llama.cpp first. Then I can't submit that PR to llama.cpp anymore?

I don't think that case is a problem at all, you authored both PRs. The question is what to do with ports of ik features to llama.cpp. Those have usually been denied. In your case the hadamard operator, I guess.

[ddh0]

Pinging @ggerganov, @JohannesGaessler, @ngxson - what do you think about these changes? I am hoping to clarify the current stance of the llama.cpp project in writing, for future contributors. Thanks.

[JohannesGaessler]

Would be fine with me.

[fairydreaming]

@ddh0 But what do you mean by a code "to originate from ik_llama.cpp"?

[ddh0]

@ddh0 But what do you mean by a code "to originate from ik_llama.cpp"?

Interpretation would be up to the maintainer / reviewer on a case-by-case basis, presumably.

[JohannesGaessler]

But what do you mean by a code "to originate from ik_llama.cpp"?

So far every time the policy was relevant it was a case of someone directly copying code from ik_llama.cpp. Those cases are completely unambiguous. And I don't think we'll have any ambiguous cases either; looking at the contributors to ik_llama.cpp there basically isn't anyone other than IK himself.

[fairydreaming]

I understand that some maintainers of llama.cpp may not want any code authored by some specific person if there's a conflict. If they are going to reject all PRs containing code authored by this person regardless of circumstances it's worth adding that to CONTRIBUTING.md as it's important information that makes life simpler for contributors.

But the part about ik_llama.cpp project puts me off and makes me confused. I have no idea what it could possibly mean - what are the specific cases this parts covers that are not already covered by the person-specific part?

[fairydreaming]

But what do you mean by a code "to originate from ik_llama.cpp"?

So far every time the policy was relevant it was a case of someone directly copying code from ik_llama.cpp. Those cases are completely unambiguous.

@JohannesGaessler I get it they were unambiguous to you - but the point of this PR is to make it unambiguous for everyone so they people don't have to waste time by researching the conflict, its current state and overall implications.

And I don't think we'll have any ambiguous cases either; looking at the contributors to ik_llama.cpp there basically isn't anyone other than IK himself.

I checked and there are several dozen people that contributed to ik_llama.cpp in the past year.

And most importantly: Happy Easter! 🥚🐣🐥🐤