llama-cpp-python/CMakeLists.txt

cmake_minimum_required(VERSION 3.21)

project(llama_cpp)

option(LLAMA_BUILD "Build llama.cpp shared library and install alongside python package" ON)
option(LLAVA_BUILD "Build llava shared library and install alongside python package" ON)

if (LLAMA_BUILD)
    set(BUILD_SHARED_LIBS "On")

    # Building llama
    if (APPLE AND NOT CMAKE_SYSTEM_PROCESSOR MATCHES "arm64")
        # Need to disable these llama.cpp flags on Apple x86_64,
        # otherwise users may encounter invalid instruction errors
        set(LLAMA_AVX "Off" CACHE BOOL "llama: enable AVX" FORCE)
        set(LLAMA_AVX2 "Off" CACHE BOOL "llama: enable AVX2" FORCE)
        set(LLAMA_FMA "Off" CACHE BOOL "llama: enable FMA" FORCE)
        set(LLAMA_F16C "Off" CACHE BOOL "llama: enable F16C" FORCE)
    endif()

    if (APPLE)
        set(LLAMA_METAL_EMBED_LIBRARY "On" CACHE BOOL "llama: embed metal library" FORCE)
    endif()

    add_subdirectory(vendor/llama.cpp)
    install(
        TARGETS llama 
        LIBRARY DESTINATION ${SKBUILD_PLATLIB_DIR}/llama_cpp
        RUNTIME DESTINATION ${SKBUILD_PLATLIB_DIR}/llama_cpp
        ARCHIVE DESTINATION ${SKBUILD_PLATLIB_DIR}/llama_cpp
        FRAMEWORK DESTINATION ${SKBUILD_PLATLIB_DIR}/llama_cpp
        RESOURCE DESTINATION ${SKBUILD_PLATLIB_DIR}/llama_cpp
    )
    # Temporary fix for https://github.com/scikit-build/scikit-build-core/issues/374
    install(
        TARGETS llama 
        LIBRARY DESTINATION ${CMAKE_CURRENT_SOURCE_DIR}/llama_cpp
        RUNTIME DESTINATION ${CMAKE_CURRENT_SOURCE_DIR}/llama_cpp
        ARCHIVE DESTINATION ${CMAKE_CURRENT_SOURCE_DIR}/llama_cpp
        FRAMEWORK DESTINATION ${CMAKE_CURRENT_SOURCE_DIR}/llama_cpp
        RESOURCE DESTINATION ${CMAKE_CURRENT_SOURCE_DIR}/llama_cpp
    )
    # Workaround for Windows + CUDA https://github.com/abetlen/llama-cpp-python/issues/563
    install(
        FILES $<TARGET_RUNTIME_DLLS:llama>
        DESTINATION ${SKBUILD_PLATLIB_DIR}/llama_cpp
    )
    install(
        FILES $<TARGET_RUNTIME_DLLS:llama>
        DESTINATION ${CMAKE_CURRENT_SOURCE_DIR}/llama_cpp
    )

    if (LLAVA_BUILD)
        if (LLAMA_CUBLAS)
            add_compile_definitions(GGML_USE_CUBLAS)
        endif()

        if (LLAMA_METAL)
            add_compile_definitions(GGML_USE_METAL)
        endif()

        # Building llava
        add_subdirectory(vendor/llama.cpp/examples/llava)
        set_target_properties(llava_shared PROPERTIES OUTPUT_NAME "llava")
        # Set CUDA_ARCHITECTURES to OFF on windows
        if (WIN32)
            set_target_properties(llava_shared PROPERTIES CUDA_ARCHITECTURES OFF)
        endif()
        install(
            TARGETS llava_shared
            LIBRARY DESTINATION ${SKBUILD_PLATLIB_DIR}/llama_cpp
            RUNTIME DESTINATION ${SKBUILD_PLATLIB_DIR}/llama_cpp
            ARCHIVE DESTINATION ${SKBUILD_PLATLIB_DIR}/llama_cpp
            FRAMEWORK DESTINATION ${SKBUILD_PLATLIB_DIR}/llama_cpp
            RESOURCE DESTINATION ${SKBUILD_PLATLIB_DIR}/llama_cpp
        )
        # Temporary fix for https://github.com/scikit-build/scikit-build-core/issues/374
        install(
            TARGETS llava_shared
            LIBRARY DESTINATION ${CMAKE_CURRENT_SOURCE_DIR}/llama_cpp
            RUNTIME DESTINATION ${CMAKE_CURRENT_SOURCE_DIR}/llama_cpp
            ARCHIVE DESTINATION ${CMAKE_CURRENT_SOURCE_DIR}/llama_cpp
            FRAMEWORK DESTINATION ${CMAKE_CURRENT_SOURCE_DIR}/llama_cpp
            RESOURCE DESTINATION ${CMAKE_CURRENT_SOURCE_DIR}/llama_cpp
        )
    endif()
endif()