Move the third-party build / deploy scripts to a separate repository

[ggerganov]

It keeps bothering me to see these scripts in the source root.
They cannot live anywhere except in the root of the repo, so therefore it is time to go.

Task: create llama.flake or llama.deploy repo and move the scripts there.

[j-f1]

Another option: move all of the source code to an src folder and leave the repo root for infrastructure things like makefiles/the README/support for packaging infrastructure.

[mdabir1203]

It keeps bothering me to see these scripts in the source root. They cannot live anywhere except in the root of the repo, so therefore it is time to go.

Task: create llama.flake or llama.deploy repo and move the scripts there.

I made a repo based on the instruction. Wanted to ask shall I include a readme ?
link: https://github.com/mdabir1203/llama.deploy/blob/main/README.md

[MostAwesomeDude]

In NixOS/nixpkgs#225058, downstream indicates that they'd prefer to see the Nix flake remain, because it ultimately gives upstream (you!) control over how everything is built and packaged. If you remove the flake, then somebody like @dit7ya will choose the packaging and it'll be added to nixpkgs. In particular, we'll probably break out ggml into its own expression, and patch it to do CPU checks at runtime. Then llama.cpp, whisper.cpp, and rwkv.cpp would each be based on that same single ggml expression.

This isn't actually desirable! Any given snapshot of llama.cpp, with its flake.nix and flake.lock, should be reproducible and have its own consistent ABI for models. This is because the flake depends on nixpkgs. But if we do it the other way around, then nixpkgs depends on whatever ABI happened to be in the given snapshot of llama.cpp. In practice, this means that users will usually have stale versions of both ggml and llama.cpp.