github/gpt4all
1
0
Fork 0
mirror of https://github.com/nomic-ai/gpt4all.git synced 2025-09-14 14:59:13 +00:00
Code Issues Packages Projects Releases Wiki Activity

support the llama.cpp CUDA backend (#2310)

Browse Source 
* rebase onto llama.cpp commit ggerganov/llama.cpp@d46dbc76f
* support for CUDA backend (enabled by default)
* partial support for Occam's Vulkan backend (disabled by default)
* partial support for HIP/ROCm backend (disabled by default)
* sync llama.cpp.cmake with upstream llama.cpp CMakeLists.txt
* changes to GPT4All backend, bindings, and chat UI to handle choice of llama.cpp backend (Kompute or CUDA)
* ship CUDA runtime with installed version
* make device selection in the UI on macOS actually do something
* model whitelist: remove dbrx, mamba, persimmon, plamo; add internlm and starcoder2

Signed-off-by: Jared Van Bortel <jared@nomic.ai>
This commit is contained in:
Jared Van Bortel
2024-05-15 15:27:50 -04:00
committed by GitHub
parent a618ca5699
commit d2a99d9bc6
22 changed files with 1360 additions and 773 deletions
Download Patch File Download Diff File Expand all files Collapse all files

68
gpt4all-chat/CMakeLists.txt
View File

@@ -17,8 +17,8 @@ if(APPLE)
endif()
set(APP_VERSION_MAJOR 2)
set(APP_VERSION_MINOR 7)
set(APP_VERSION_PATCH 6)
set(APP_VERSION_MINOR 8)
set(APP_VERSION_PATCH 0)
set(APP_VERSION "${APP_VERSION_MAJOR}.${APP_VERSION_MINOR}.${APP_VERSION_PATCH}")
# Include the binary directory for the generated header file
@@ -65,7 +65,7 @@ add_subdirectory(../gpt4all-backend llmodel)
set(METAL_SHADER_FILE)
if(${CMAKE_SYSTEM_NAME} MATCHES Darwin)
set(METAL_SHADER_FILE ../gpt4all-backend/llama.cpp-mainline/ggml-metal.metal)
set(METAL_SHADER_FILE ${CMAKE_RUNTIME_OUTPUT_DIRECTORY}/default.metallib)
endif()
set(APP_ICON_RESOURCE)
@@ -185,7 +185,6 @@ if(METAL_SHADER_FILE)
set_target_properties(chat PROPERTIES
RESOURCE ${METAL_SHADER_FILE}
)
configure_file(${METAL_SHADER_FILE} bin/ggml-metal.metal COPYONLY)
endif()
target_compile_definitions(chat
@@ -207,18 +206,61 @@ if(CMAKE_INSTALL_PREFIX_INITIALIZED_TO_DEFAULT)
endif()
install(TARGETS chat DESTINATION bin COMPONENT ${COMPONENT_NAME_MAIN})
install(TARGETS llmodel DESTINATION lib COMPONENT ${COMPONENT_NAME_MAIN})
install(
TARGETS llmodel
LIBRARY DESTINATION lib COMPONENT ${COMPONENT_NAME_MAIN} # .so/.dylib
RUNTIME DESTINATION bin COMPONENT ${COMPONENT_NAME_MAIN} # .dll
)
# We should probably iterate through the list of the cmake for backend, but these need to be installed
# to the this component's dir for the finicky qt installer to work
install(TARGETS gptj-avxonly DESTINATION lib COMPONENT ${COMPONENT_NAME_MAIN})
install(TARGETS gptj-default DESTINATION lib COMPONENT ${COMPONENT_NAME_MAIN})
install(TARGETS llama-mainline-avxonly DESTINATION lib COMPONENT ${COMPONENT_NAME_MAIN})
install(TARGETS llama-mainline-default DESTINATION lib COMPONENT ${COMPONENT_NAME_MAIN})
install(TARGETS llamamodel-mainline-avxonly DESTINATION lib COMPONENT ${COMPONENT_NAME_MAIN})
install(TARGETS llamamodel-mainline-default DESTINATION lib COMPONENT ${COMPONENT_NAME_MAIN})
if(APPLE)
install(TARGETS llamamodel-mainline-metal DESTINATION lib COMPONENT ${COMPONENT_NAME_MAIN})
if (LLMODEL_KOMPUTE)
set(MODEL_IMPL_TARGETS
llamamodel-mainline-kompute
llamamodel-mainline-kompute-avxonly
gptj-kompute
gptj-kompute-avxonly
)
else()
set(MODEL_IMPL_TARGETS
llamamodel-mainline-cpu
llamamodel-mainline-cpu-avxonly
gptj-cpu
gptj-cpu-avxonly
)
endif()
if (APPLE)
list(APPEND MODEL_IMPL_TARGETS llamamodel-mainline-metal)
endif()
install(
TARGETS ${MODEL_IMPL_TARGETS}
LIBRARY DESTINATION lib COMPONENT ${COMPONENT_NAME_MAIN} # .so/.dylib
RUNTIME DESTINATION lib COMPONENT ${COMPONENT_NAME_MAIN} # .dll
)
if (LLMODEL_CUDA)
set_property(TARGET llamamodel-mainline-cuda llamamodel-mainline-cuda-avxonly
APPEND PROPERTY INSTALL_RPATH "$ORIGIN")
install(
TARGETS llamamodel-mainline-cuda
llamamodel-mainline-cuda-avxonly
RUNTIME_DEPENDENCY_SET llama-cuda-deps
LIBRARY DESTINATION lib COMPONENT ${COMPONENT_NAME_MAIN} # .so/.dylib
RUNTIME DESTINATION lib COMPONENT ${COMPONENT_NAME_MAIN} # .dll
)
if (WIN32)
install(
RUNTIME_DEPENDENCY_SET llama-cuda-deps
PRE_EXCLUDE_REGEXES "^(nvcuda|api-ms-.*)\\.dll$"
POST_INCLUDE_REGEXES "(^|[/\\\\])(lib)?(cuda|cublas)" POST_EXCLUDE_REGEXES .
DIRECTORIES "${CUDAToolkit_BIN_DIR}"
DESTINATION lib COMPONENT ${COMPONENT_NAME_MAIN}
)
endif()
endif()
set(CPACK_GENERATOR "IFW")