ExternalVendorCode/LocalAI
1
0
Fork 0
mirror of https://github.com/mudler/LocalAI.git synced 2024-12-18 12:26:26 +00:00
Code Issues Actions 12 Packages Projects Releases Wiki Activity

feat: unify and propagate CMAKE_ARGS to GGML-based backends

Browse Source 
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
This commit is contained in:
Ettore Di Giacinto 2024-12-11 22:01:05 +01:00
parent c85f46a71d
commit 894a30296a
2 changed files with 45 additions and 42 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

49
Makefile
View File

@ -40,6 +40,7 @@ ONNX_OS?=linux
export BUILD_TYPE?=
export STABLE_BUILD_TYPE?=$(BUILD_TYPE)
export GGML_CMAKE_ARGS?=
export CMAKE_ARGS?=
export BACKEND_LIBS?=
@ -88,9 +89,45 @@ ifndef UNAME_S
UNAME_S := $(shell uname -s)
endif
# IF native is false, we add -DGGML_NATIVE=OFF to CMAKE_ARGS
# IF native is false, we add -DGGML_NATIVE=OFF to GGML_CMAKE_ARGS
ifeq ($(NATIVE),false)
CMAKE_ARGS+=-DGGML_NATIVE=OFF
GGML_CMAKE_ARGS+=-DGGML_NATIVE=OFF
endif
# Disable Shared libs as we are linking on static gRPC and we can't mix shared and static
GGML_CMAKE_ARGS+=-DBUILD_SHARED_LIBS=OFF
# If build type is cublas, then we set -DGGML_CUDA=ON to GGML_CMAKE_ARGS automatically
ifeq ($(BUILD_TYPE),cublas)
GGML_CMAKE_ARGS+=-DGGML_CUDA=ON
# If build type is openblas then we set -DGGML_BLAS=ON -DGGML_BLAS_VENDOR=OpenBLAS
# to GGML_CMAKE_ARGS automatically
else ifeq ($(BUILD_TYPE),openblas)
GGML_CMAKE_ARGS+=-DGGML_BLAS=ON -DGGML_BLAS_VENDOR=OpenBLAS
# If build type is clblas (openCL) we set -DGGML_CLBLAST=ON -DCLBlast_DIR=/some/path
else ifeq ($(BUILD_TYPE),clblas)
GGML_CMAKE_ARGS+=-DGGML_CLBLAST=ON -DCLBlast_DIR=/some/path
# If it's hipblas we do have also to set CC=/opt/rocm/llvm/bin/clang CXX=/opt/rocm/llvm/bin/clang++
else ifeq ($(BUILD_TYPE),hipblas)
GGML_CMAKE_ARGS+=-DGGML_HIP=ON
# If it's OSX, DO NOT embed the metal library - -DGGML_METAL_EMBED_LIBRARY=ON requires further investigation
# But if it's OSX without metal, disable it here
else ifeq ($(OS),Darwin)
ifneq ($(BUILD_TYPE),metal)
GGML_CMAKE_ARGS+=-DGGML_METAL=OFF
else
GGML_CMAKE_ARGS+=-DGGML_METAL=ON
GGML_CMAKE_ARGS+=-DGGML_METAL_EMBED_LIBRARY=ON
TARGET+=--target ggml-metal
endif
endif
ifeq ($(BUILD_TYPE),sycl_f16)
GGML_CMAKE_ARGS+=-DGGML_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx -DGGML_SYCL_F16=ON
endif
ifeq ($(BUILD_TYPE),sycl_f32)
GGML_CMAKE_ARGS+=-DGGML_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx
endif
# Detect if we are running on arm64
@ -117,7 +154,7 @@ ifeq ($(OS),Darwin)
BUILD_TYPE=metal
# disable metal if on Darwin and any other value is explicitly passed.
else ifneq ($(BUILD_TYPE),metal)
CMAKE_ARGS+=-DGGML_METAL=OFF
GGML_CMAKE_ARGS+=-DGGML_METAL=OFF
export GGML_NO_ACCELERATE=1
export GGML_NO_METAL=1
endif
@ -142,7 +179,7 @@ ifeq ($(BUILD_TYPE),cublas)
endif
ifeq ($(BUILD_TYPE),vulkan)
CMAKE_ARGS+=-DGGML_VULKAN=1
GGML_CMAKE_ARGS+=-DGGML_VULKAN=1
endif
ifneq (,$(findstring sycl,$(BUILD_TYPE)))
@ -164,7 +201,7 @@ ifeq ($(BUILD_TYPE),hipblas)
export GGML_HIP=1
GPU_TARGETS ?= gfx900,gfx906,gfx908,gfx940,gfx941,gfx942,gfx90a,gfx1030,gfx1031,gfx1100,gfx1101
AMDGPU_TARGETS ?= "$(GPU_TARGETS)"
CMAKE_ARGS+=-DGGML_HIP=ON -DAMDGPU_TARGETS="$(AMDGPU_TARGETS)" -DGPU_TARGETS="$(GPU_TARGETS)"
GGML_CMAKE_ARGS+=-DGGML_HIP=ON -DAMDGPU_TARGETS="$(AMDGPU_TARGETS)" -DGPU_TARGETS="$(GPU_TARGETS)"
CGO_LDFLAGS += -O3 --rtlib=compiler-rt -unwindlib=libgcc -lhipblas -lrocblas --hip-link -L${ROCM_HOME}/lib/llvm/lib
endif
@ -235,6 +272,8 @@ ifeq ($(BUILD_API_ONLY),true)
GRPC_BACKENDS=
endif
export CMAKE_ARGS?=$(GGML_CMAKE_ARGS)
.PHONY: all test build vendor get-sources prepare-sources prepare
all: help

38
backend/cpp/llama/Makefile
View File

@ -7,42 +7,6 @@ BUILD_TYPE?=
ONEAPI_VARS?=/opt/intel/oneapi/setvars.sh
TARGET?=--target grpc-server
# Disable Shared libs as we are linking on static gRPC and we can't mix shared and static
CMAKE_ARGS+=-DBUILD_SHARED_LIBS=OFF
# If build type is cublas, then we set -DGGML_CUDA=ON to CMAKE_ARGS automatically
ifeq ($(BUILD_TYPE),cublas)
CMAKE_ARGS+=-DGGML_CUDA=ON
# If build type is openblas then we set -DGGML_BLAS=ON -DGGML_BLAS_VENDOR=OpenBLAS
# to CMAKE_ARGS automatically
else ifeq ($(BUILD_TYPE),openblas)
CMAKE_ARGS+=-DGGML_BLAS=ON -DGGML_BLAS_VENDOR=OpenBLAS
# If build type is clblas (openCL) we set -DGGML_CLBLAST=ON -DCLBlast_DIR=/some/path
else ifeq ($(BUILD_TYPE),clblas)
CMAKE_ARGS+=-DGGML_CLBLAST=ON -DCLBlast_DIR=/some/path
# If it's hipblas we do have also to set CC=/opt/rocm/llvm/bin/clang CXX=/opt/rocm/llvm/bin/clang++
else ifeq ($(BUILD_TYPE),hipblas)
CMAKE_ARGS+=-DGGML_HIP=ON
# If it's OSX, DO NOT embed the metal library - -DGGML_METAL_EMBED_LIBRARY=ON requires further investigation
# But if it's OSX without metal, disable it here
else ifeq ($(OS),Darwin)
ifneq ($(BUILD_TYPE),metal)
CMAKE_ARGS+=-DGGML_METAL=OFF
else
CMAKE_ARGS+=-DGGML_METAL=ON
CMAKE_ARGS+=-DGGML_METAL_EMBED_LIBRARY=ON
TARGET+=--target ggml-metal
endif
endif
ifeq ($(BUILD_TYPE),sycl_f16)
CMAKE_ARGS+=-DGGML_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx -DGGML_SYCL_F16=ON
endif
ifeq ($(BUILD_TYPE),sycl_f32)
CMAKE_ARGS+=-DGGML_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx
endif
llama.cpp:
mkdir -p llama.cpp
cd llama.cpp && \
@ -70,7 +34,7 @@ clean: purge
rm -rf llama.cpp
grpc-server: llama.cpp llama.cpp/examples/grpc-server
@echo "Building grpc-server with $(BUILD_TYPE) build type and $(CMAKE_ARGS)"
@echo "Building grpc-server for llama.cpp with $(BUILD_TYPE) build type and $(CMAKE_ARGS)"
ifneq (,$(findstring sycl,$(BUILD_TYPE)))
+bash -c "source $(ONEAPI_VARS); \
cd llama.cpp && mkdir -p build && cd build && cmake .. $(CMAKE_ARGS) && cmake --build . --config Release $(TARGET)"