Skip to content

Commit

Permalink
fix: merge with code
Browse files Browse the repository at this point in the history
  • Loading branch information
@JoanFM
JoanFM committed Jun 5, 2024
2 parents 8a359c4 + 7ab6023 commit 7b26708
Show file tree
Hide file tree
Showing 213 changed files with 110,092 additions and 47,333 deletions.
46 changes: 12 additions & 34 deletions .github/workflows/build.yml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -294,12 +294,22 @@ jobs:
- name: Build
id: cmake_build
if: ${{ matrix.sanitizer != 'THREAD' }}
run: |
mkdir build
cd build
cmake .. -DLLAMA_FATAL_WARNINGS=ON -DLLAMA_SANITIZE_${{ matrix.sanitizer }}=ON -DCMAKE_BUILD_TYPE=${{ matrix.build_type }}
cmake --build . --config ${{ matrix.build_type }} -j $(nproc)
- name: Build (no OpenMP)
id: cmake_build_no_openmp
if: ${{ matrix.sanitizer == 'THREAD' }}
run: |
mkdir build
cd build
cmake .. -DLLAMA_FATAL_WARNINGS=ON -DLLAMA_SANITIZE_${{ matrix.sanitizer }}=ON -DCMAKE_BUILD_TYPE=${{ matrix.build_type }} -DLLAMA_OPENMP=OFF
cmake --build . --config ${{ matrix.build_type }} -j $(nproc)
- name: Test
id: cmake_test
run: |
Expand Down Expand Up @@ -678,8 +688,6 @@ jobs:

env:
OPENBLAS_VERSION: 0.3.23
OPENCL_VERSION: 2023.04.17
CLBLAST_VERSION: 1.6.0
SDE_VERSION: 9.33.0-2024-01-07
VULKAN_VERSION: 1.3.261.1

Expand All @@ -696,8 +704,6 @@ jobs:
defines: '-DLLAMA_NATIVE=OFF -DLLAMA_BUILD_SERVER=ON -DLLAMA_AVX2=OFF -DBUILD_SHARED_LIBS=ON'
- build: 'avx512-x64'
defines: '-DLLAMA_NATIVE=OFF -DLLAMA_BUILD_SERVER=ON -DLLAMA_AVX512=ON -DBUILD_SHARED_LIBS=ON'
- build: 'clblast-x64'
defines: '-DLLAMA_NATIVE=OFF -DLLAMA_BUILD_SERVER=ON -DLLAMA_CLBLAST=ON -DBUILD_SHARED_LIBS=ON -DCMAKE_PREFIX_PATH="$env:RUNNER_TEMP/clblast"'
- build: 'openblas-x64'
defines: '-DLLAMA_NATIVE=OFF -DLLAMA_BUILD_SERVER=ON -DLLAMA_BLAS=ON -DBUILD_SHARED_LIBS=ON -DLLAMA_BLAS_VENDOR=OpenBLAS -DBLAS_INCLUDE_DIRS="$env:RUNNER_TEMP/openblas/include" -DBLAS_LIBRARIES="$env:RUNNER_TEMP/openblas/lib/openblas.lib"'
- build: 'kompute-x64'
Expand All @@ -722,27 +728,6 @@ jobs:
run: |
git submodule update --init kompute
- name: Download OpenCL SDK
id: get_opencl
if: ${{ matrix.build == 'clblast-x64' }}
run: |
curl.exe -o $env:RUNNER_TEMP/opencl.zip -L "https://github.com/KhronosGroup/OpenCL-SDK/releases/download/v${env:OPENCL_VERSION}/OpenCL-SDK-v${env:OPENCL_VERSION}-Win-x64.zip"
mkdir $env:RUNNER_TEMP/opencl
tar.exe -xvf $env:RUNNER_TEMP/opencl.zip --strip-components=1 -C $env:RUNNER_TEMP/opencl
- name: Download CLBlast
id: get_clblast
if: ${{ matrix.build == 'clblast-x64' }}
run: |
curl.exe -o $env:RUNNER_TEMP/clblast.7z -L "https://github.com/CNugteren/CLBlast/releases/download/${env:CLBLAST_VERSION}/CLBlast-${env:CLBLAST_VERSION}-windows-x64.7z"
curl.exe -o $env:RUNNER_TEMP/CLBlast.LICENSE.txt -L "https://github.com/CNugteren/CLBlast/raw/${env:CLBLAST_VERSION}/LICENSE"
7z x "-o${env:RUNNER_TEMP}" $env:RUNNER_TEMP/clblast.7z
rename-item $env:RUNNER_TEMP/CLBlast-${env:CLBLAST_VERSION}-windows-x64 clblast
foreach ($f in (gci -Recurse -Path "$env:RUNNER_TEMP/clblast" -Filter '*.cmake')) {
$txt = Get-Content -Path $f -Raw
$txt.Replace('C:/vcpkg/packages/opencl_x64-windows/', "$($env:RUNNER_TEMP.Replace('\','/'))/opencl/") | Set-Content -Path $f -Encoding UTF8
}
- name: Download OpenBLAS
id: get_openblas
if: ${{ matrix.build == 'openblas-x64' }}
Expand Down Expand Up @@ -776,13 +761,6 @@ jobs:
cmake -S . -B build ${{ matrix.defines }}
cmake --build build --config Release -j ${env:NUMBER_OF_PROCESSORS}
- name: Add clblast.dll
id: add_clblast_dll
if: ${{ matrix.build == 'clblast-x64' }}
run: |
cp $env:RUNNER_TEMP/clblast/lib/clblast.dll ./build/bin/Release
cp $env:RUNNER_TEMP/CLBlast.LICENSE.txt ./build/bin/Release/CLBlast-${env:CLBLAST_VERSION}.txt
- name: Add libopenblas.dll
id: add_libopenblas_dll
if: ${{ matrix.build == 'openblas-x64' }}
Expand All @@ -806,7 +784,7 @@ jobs:
- name: Test
id: cmake_test
# not all machines have native AVX-512
if: ${{ matrix.build != 'msvc-arm64' && matrix.build != 'llvm-arm64' && matrix.build != 'clblast-x64' && matrix.build != 'kompute-x64' && matrix.build != 'vulkan-x64' && (matrix.build != 'avx512-x64' || env.HAS_AVX512F == '1') }}
if: ${{ matrix.build != 'msvc-arm64' && matrix.build != 'llvm-arm64' && matrix.build != 'kompute-x64' && matrix.build != 'vulkan-x64' && (matrix.build != 'avx512-x64' || env.HAS_AVX512F == '1') }}
run: |
cd build
ctest -L main -C Release --verbose --timeout 900
Expand Down Expand Up @@ -1061,7 +1039,7 @@ jobs:
# hypervisor: 'qemu'
# run: |
# sudo pkg update
# sudo pkg install -y gmake automake autoconf pkgconf llvm15 clinfo clover opencl clblast openblas
# sudo pkg install -y gmake automake autoconf pkgconf llvm15 openblas
# gmake CC=/usr/local/bin/clang15 CXX=/usr/local/bin/clang++15 -j `sysctl -n hw.ncpu`

release:
Expand Down
3 changes: 3 additions & 0 deletions .gitignore
View file
Original file line number Diff line number Diff line change
Expand Up @@ -34,9 +34,11 @@ ggml-metal-embed.metal
lcov-report/
gcovr-report/

tags
build*
!build.zig
cmake-build-*
android-ndk-*
out/
tmp/

Expand Down Expand Up @@ -105,6 +107,7 @@ examples/jeopardy/results.txt
examples/server/*.html.hpp
examples/server/*.js.hpp
examples/server/*.mjs.hpp
examples/server/*.css.hpp

poetry.lock
poetry.toml
Expand Down
85 changes: 62 additions & 23 deletions CMakeLists.txt
View file
Original file line number Diff line number Diff line change
Expand Up @@ -106,11 +106,11 @@ set(LLAMA_CUDA_PEER_MAX_BATCH_SIZE "128" CACHE STRING
"llama: max. batch size for using peer access")
option(LLAMA_CUDA_NO_PEER_COPY "llama: do not use peer to peer copies" OFF)
option(LLAMA_CUDA_NO_VMM "llama: do not try to use CUDA VMM" OFF)
option(LLAMA_CUDA_FA_ALL_QUANTS "llama: compile all quants for FlashAttention" OFF)

option(LLAMA_CURL "llama: use libcurl to download model from an URL" OFF)
option(LLAMA_HIPBLAS "llama: use hipBLAS" OFF)
option(LLAMA_HIP_UMA "llama: use HIP unified memory architecture" OFF)
option(LLAMA_CLBLAST "llama: use CLBlast" OFF)
option(LLAMA_VULKAN "llama: use Vulkan" OFF)
option(LLAMA_VULKAN_CHECK_RESULTS "llama: run Vulkan op checks" OFF)
option(LLAMA_VULKAN_DEBUG "llama: enable Vulkan debug output" OFF)
Expand All @@ -125,6 +125,7 @@ set(LLAMA_METAL_MACOSX_VERSION_MIN "" CACHE STRING
set(LLAMA_METAL_STD "" CACHE STRING "llama: metal standard version (-std flag)")
option(LLAMA_KOMPUTE "llama: use Kompute" OFF)
option(LLAMA_RPC "llama: use RPC" OFF)
option(LLAMA_OPENMP "llama: use OpenMP" ON)
option(LLAMA_SYCL "llama: use SYCL" OFF)
option(LLAMA_SYCL_F16 "llama: use 16 bit floats for sycl calculations" OFF)
set(LLAMA_SYCL_TARGET "INTEL" CACHE STRING "llama: sycl target device")
Expand Down Expand Up @@ -295,6 +296,17 @@ if (LLAMA_METAL)
)
endif()

if (LLAMA_OPENMP)
find_package(OpenMP)
if (OpenMP_FOUND)
message(STATUS "OpenMP found")
add_compile_definitions(GGML_USE_OPENMP)
set(LLAMA_EXTRA_LIBS ${LLAMA_EXTRA_LIBS} OpenMP::OpenMP_C OpenMP::OpenMP_CXX)
else()
message(WARNING "OpenMP not found")
endif()
endif()

if (LLAMA_BLAS)
if (LLAMA_STATIC)
set(BLA_STATIC ON)
Expand Down Expand Up @@ -402,6 +414,8 @@ if (LLAMA_CUDA)

file(GLOB GGML_SOURCES_CUDA "ggml-cuda/*.cu")
list(APPEND GGML_SOURCES_CUDA "ggml-cuda.cu")
file(GLOB SRCS "ggml-cuda/template-instances/fattn-wmma*.cu")
list(APPEND GGML_SOURCES_CUDA ${SRCS})

add_compile_definitions(GGML_USE_CUDA)
add_compile_definitions(GGML_CUDA_USE_GRAPHS)
Expand All @@ -427,6 +441,18 @@ if (LLAMA_CUDA)
if (LLAMA_CUDA_NO_PEER_COPY)
add_compile_definitions(GGML_CUDA_NO_PEER_COPY)
endif()
if (LLAMA_CUDA_FA_ALL_QUANTS)
file(GLOB SRCS "ggml-cuda/template-instances/fattn-vec*.cu")
list(APPEND GGML_SOURCES_CUDA ${SRCS})
add_compile_definitions(GGML_CUDA_FA_ALL_QUANTS)
else()
file(GLOB SRCS "ggml-cuda/template-instances/fattn-vec*q4_0-q4_0.cu")
list(APPEND GGML_SOURCES_CUDA ${SRCS})
file(GLOB SRCS "ggml-cuda/template-instances/fattn-vec*q8_0-q8_0.cu")
list(APPEND GGML_SOURCES_CUDA ${SRCS})
file(GLOB SRCS "ggml-cuda/template-instances/fattn-vec*f16-f16.cu")
list(APPEND GGML_SOURCES_CUDA ${SRCS})
endif()

if (LLAMA_STATIC)
if (WIN32)
Expand Down Expand Up @@ -475,22 +501,6 @@ if (LLAMA_RPC)
set(GGML_SOURCES_RPC ggml-rpc.cpp)
endif()

if (LLAMA_CLBLAST)
find_package(CLBlast)
if (CLBlast_FOUND)
message(STATUS "CLBlast found")

set(GGML_HEADERS_OPENCL ggml-opencl.h)
set(GGML_SOURCES_OPENCL ggml-opencl.cpp)

add_compile_definitions(GGML_USE_CLBLAST)

set(LLAMA_EXTRA_LIBS ${LLAMA_EXTRA_LIBS} clblast)
else()
message(WARNING "CLBlast not found")
endif()
endif()

if (LLAMA_VULKAN)
find_package(Vulkan)
if (Vulkan_FOUND)
Expand Down Expand Up @@ -530,12 +540,17 @@ if (LLAMA_VULKAN)
endif()

if (LLAMA_HIPBLAS)
if ($ENV{ROCM_PATH})
set(ROCM_PATH $ENV{ROCM_PATH})
if (NOT EXISTS $ENV{ROCM_PATH})
if (NOT EXISTS /opt/rocm)
set(ROCM_PATH /usr)
else()
set(ROCM_PATH /opt/rocm)
endif()
else()
set(ROCM_PATH /opt/rocm)
set(ROCM_PATH $ENV{ROCM_PATH})
endif()
list(APPEND CMAKE_PREFIX_PATH ${ROCM_PATH})
list(APPEND CMAKE_PREFIX_PATH "${ROCM_PATH}/lib64/cmake")

# CMake on Windows doesn't support the HIP language yet
if(WIN32)
Expand Down Expand Up @@ -571,6 +586,8 @@ if (LLAMA_HIPBLAS)

file(GLOB GGML_SOURCES_ROCM "ggml-cuda/*.cu")
list(APPEND GGML_SOURCES_ROCM "ggml-cuda.cu")
file(GLOB SRCS "ggml-cuda/template-instances/fattn-wmma*.cu")
list(APPEND GGML_SOURCES_ROCM ${SRCS})

add_compile_definitions(GGML_USE_HIPBLAS GGML_USE_CUDA)

Expand All @@ -590,6 +607,19 @@ if (LLAMA_HIPBLAS)
add_compile_definitions(GGML_CUDA_NO_PEER_COPY)
endif()

if (LLAMA_CUDA_FA_ALL_QUANTS)
file(GLOB SRCS "ggml-cuda/template-instances/fattn-vec*.cu")
list(APPEND GGML_SOURCES_ROCM ${SRCS})
add_compile_definitions(GGML_CUDA_FA_ALL_QUANTS)
else()
file(GLOB SRCS "ggml-cuda/template-instances/fattn-vec*q4_0-q4_0.cu")
list(APPEND GGML_SOURCES_ROCM ${SRCS})
file(GLOB SRCS "ggml-cuda/template-instances/fattn-vec*q8_0-q8_0.cu")
list(APPEND GGML_SOURCES_ROCM ${SRCS})
file(GLOB SRCS "ggml-cuda/template-instances/fattn-vec*f16-f16.cu")
list(APPEND GGML_SOURCES_ROCM ${SRCS})
endif()

add_compile_definitions(GGML_CUDA_DMMV_X=${LLAMA_CUDA_DMMV_X})
add_compile_definitions(GGML_CUDA_MMV_Y=${LLAMA_CUDA_MMV_Y})
add_compile_definitions(K_QUANTS_PER_ITERATION=${LLAMA_CUDA_KQUANTS_ITER})
Expand Down Expand Up @@ -747,6 +777,7 @@ if (LLAMA_KOMPUTE)
kompute-shaders/op_mul_mat_q4_0.comp
kompute-shaders/op_mul_mat_q4_1.comp
kompute-shaders/op_mul_mat_q6_k.comp
kompute-shaders/op_getrows_f32.comp
kompute-shaders/op_getrows_f16.comp
kompute-shaders/op_getrows_q4_0.comp
kompute-shaders/op_getrows_q4_1.comp
Expand Down Expand Up @@ -779,6 +810,7 @@ if (LLAMA_KOMPUTE)
shaderop_mul_mat_q4_0.h
shaderop_mul_mat_q4_1.h
shaderop_mul_mat_q6_k.h
shaderop_getrows_f32.h
shaderop_getrows_f16.h
shaderop_getrows_q4_0.h
shaderop_getrows_q4_1.h
Expand Down Expand Up @@ -1216,7 +1248,6 @@ add_library(ggml OBJECT
ggml-quants.c
ggml-quants.h
${GGML_SOURCES_CUDA} ${GGML_HEADERS_CUDA}
${GGML_SOURCES_OPENCL} ${GGML_HEADERS_OPENCL}
${GGML_SOURCES_METAL} ${GGML_HEADERS_METAL}
${GGML_SOURCES_RPC} ${GGML_HEADERS_RPC}
${GGML_SOURCES_EXTRA} ${GGML_HEADERS_EXTRA}
Expand Down Expand Up @@ -1304,8 +1335,9 @@ install(FILES ${CMAKE_CURRENT_BINARY_DIR}/LlamaConfig.cmake
DESTINATION ${CMAKE_INSTALL_LIBDIR}/cmake/Llama)

set(GGML_PUBLIC_HEADERS "ggml.h" "ggml-alloc.h" "ggml-backend.h"
"${GGML_HEADERS_CUDA}" "${GGML_HEADERS_OPENCL}"
"${GGML_HEADERS_METAL}" "${GGML_HEADERS_EXTRA}")
"${GGML_HEADERS_CUDA}"
"${GGML_HEADERS_METAL}"
"${GGML_HEADERS_EXTRA}")

set_target_properties(ggml PROPERTIES PUBLIC_HEADER "${GGML_PUBLIC_HEADERS}")
install(TARGETS ggml PUBLIC_HEADER)
Expand Down Expand Up @@ -1341,6 +1373,13 @@ if (LLAMA_METAL)
endif()
endif()

configure_file(cmake/llama.pc.in
"${CMAKE_CURRENT_BINARY_DIR}/llama.pc"
@ONLY)

install(FILES "${CMAKE_CURRENT_BINARY_DIR}/llama.pc"
DESTINATION lib/pkgconfig)

#
# programs, examples and tests
#
Expand Down
Loading

0 comments on commit 7b26708

Please sign in to comment.