The build system is based on CMake; for details, refer to CMakePresets.json. Typically you will want to use the release_template_debug and release_template_release presets.
$env:VCPKG_OVERLAY_PORTS="vcpkg/ports"
$env:VCPKG_DEFAULT_TRIPLET="x64-windows"
vcpkg install ggml[blas,cuda] llama-cpp[server,tools] whisper-cpp[cuda]References:
- https://github.com/ggml-org/llama.cpp/blob/master/docs/build.md#override-compute-capability-specifications
- CMAKE_CUDA_ARCHITECTURES https://github.com/ggml-org/llama.cpp/blob/4f0e43da6f8f6e9390d88409610098ec2d2dc5c7/ggml/src/ggml-cuda/CMakeLists.txt#L8
# configure
cmake --preset release_template_debug
# build
cmake --build build/release_template_debug --config RelWithDebInfo# Inside the `project` folder:
godot --doctool ../ --gdextension-docs
# The command above generates XML files inside the `doc_classes` folder.