Open mdegans opened 1 month ago
Next step is Windows + amd64 + CUDA. I do have a Microsoft arm64 dev box which I believe has an NPU. I'm unsure if llama.cpp
supports it or if there's a backend that does.
That's something for later. Very few people have such a setup, although it will likely become more popular with the new "Copilot" PCs that require an NPU and might be Arm64.
Nothing should actually prevent a windows build. It's more matter of using the right compiler, dependencies, and setting up CI. MSVC does not build out of the box. This is likely an issue with
drama_llama
and thellama-cpp-sys-3
bindings as well.