Does llama.cpp support CUDA through HIP?

[MarioIshac]

I am already running llama.cpp using HIP for ROCm and it is great, but had the above question out of curiosity. If HIP is a thin wrapper of CUDA on CUDA machines, does that mean that following this AMD guide would produce a llama-server / artifact that would be GPU accelerated on an Nvidia machine?

[IMbackK]

sure yeah that should work, i dont think anyone has tried however