Skip to content

Commit 9f2da58

Browse files
slarenslaren
authored
llama : build windows releases with dl backends (#13220)
1 parent 93c4e23 commit 9f2da58

File tree

4 files changed

+30
-58
lines changed

4 files changed

+30
-58
lines changed

‎.github/workflows/build.yml‎

Lines changed: 26 additions & 45 deletions
Original file line numberDiff line numberDiff line change
@@ -771,7 +771,7 @@ jobs:
771771
uses: hendrikmuhs/[email protected]
772772
with:
773773
key: windows-msys2
774-
variant: sccache
774+
variant: ccache
775775
evict-old-files: 1d
776776

777777
- name: Setup ${{ matrix.sys }}
@@ -814,26 +814,18 @@ jobs:
814814
strategy:
815815
matrix:
816816
include:
817-
- build: 'noavx-x64'
818-
defines: '-DGGML_NATIVE=OFF -DLLAMA_BUILD_SERVER=ON -DGGML_RPC=ON -DGGML_AVX=OFF -DGGML_AVX2=OFF -DGGML_FMA=OFF'
819-
- build: 'avx2-x64'
820-
defines: '-DGGML_NATIVE=OFF -DLLAMA_BUILD_SERVER=ON -DGGML_RPC=ON'
821-
- build: 'avx-x64'
822-
defines: '-DGGML_NATIVE=OFF -DLLAMA_BUILD_SERVER=ON -DGGML_RPC=ON -DGGML_AVX2=OFF'
823-
- build: 'avx512-x64'
824-
defines: '-DGGML_NATIVE=OFF -DLLAMA_BUILD_SERVER=ON -DGGML_RPC=ON -DGGML_AVX512=ON'
817+
- build: 'cpu-x64'
818+
defines: '-G "Ninja Multi-Config" -D CMAKE_TOOLCHAIN_FILE=cmake/x64-windows-llvm.cmake -DGGML_NATIVE=OFF -DLLAMA_BUILD_SERVER=ON -DGGML_RPC=ON -DGGML_BACKEND_DL=ON -DGGML_CPU_ALL_VARIANTS=ON -DGGML_OPENMP=OFF'
825819
- build: 'openblas-x64'
826-
defines: '-DGGML_NATIVE=OFF -DLLAMA_BUILD_SERVER=ON -DGGML_RPC=ON -DGGML_BLAS=ON -DGGML_BLAS_VENDOR=OpenBLAS -DBLAS_INCLUDE_DIRS="$env:RUNNER_TEMP/openblas/include" -DBLAS_LIBRARIES="$env:RUNNER_TEMP/openblas/lib/openblas.lib"'
827-
- build: 'kompute-x64'
828-
defines: '-DGGML_NATIVE=OFF -DLLAMA_BUILD_SERVER=ON -DGGML_RPC=ON -DGGML_KOMPUTE=ON -DKOMPUTE_OPT_DISABLE_VULKAN_VERSION_CHECK=ON'
820+
defines: '-G "Ninja Multi-Config" -D CMAKE_TOOLCHAIN_FILE=cmake/x64-windows-llvm.cmake -DGGML_NATIVE=OFF -DLLAMA_BUILD_SERVER=ON -DGGML_RPC=ON -DGGML_BACKEND_DL=ON -DGGML_CPU_ALL_VARIANTS=ON -DGGML_OPENMP=OFF -DGGML_BLAS=ON -DGGML_BLAS_VENDOR=OpenBLAS -DBLAS_INCLUDE_DIRS="$env:RUNNER_TEMP/openblas/include" -DBLAS_LIBRARIES="$env:RUNNER_TEMP/openblas/lib/openblas.lib"'
829821
- build: 'vulkan-x64'
830-
defines: '-DGGML_NATIVE=OFF -DLLAMA_BUILD_SERVER=ON -DGGML_RPC=ON -DGGML_VULKAN=ON'
822+
defines: '-DGGML_NATIVE=OFF -DLLAMA_BUILD_SERVER=ON -DGGML_RPC=ON -DGGML_BACKEND_DL=ON -DGGML_CPU_ALL_VARIANTS=ON -DGGML_VULKAN=ON'
831823
- build: 'llvm-arm64'
832824
defines: '-G "Ninja Multi-Config" -D CMAKE_TOOLCHAIN_FILE=cmake/arm64-windows-llvm.cmake -DGGML_NATIVE=OFF -DLLAMA_BUILD_SERVER=ON'
833-
- build: 'msvc-arm64'
834-
defines: '-G "Ninja Multi-Config" -D CMAKE_TOOLCHAIN_FILE=cmake/arm64-windows-msvc.cmake -DGGML_NATIVE=OFF -DLLAMA_BUILD_SERVER=ON'
835825
- build: 'llvm-arm64-opencl-adreno'
836826
defines: '-G "Ninja Multi-Config" -D CMAKE_TOOLCHAIN_FILE=cmake/arm64-windows-llvm.cmake -DCMAKE_PREFIX_PATH="$env:RUNNER_TEMP/opencl-arm64-release" -DGGML_OPENCL=ON -DGGML_OPENCL_USE_ADRENO_KERNELS=ON'
827+
# - build: 'kompute-x64'
828+
# defines: '-G "Ninja Multi-Config" -D CMAKE_TOOLCHAIN_FILE=cmake/x64-windows-llvm.cmake -DGGML_NATIVE=OFF -DLLAMA_BUILD_SERVER=ON -DGGML_RPC=ON -DGGML_BACKEND_DL=ON -DGGML_CPU_ALL_VARIANTS=ON -DGGML_OPENMP=OFF -DGGML_KOMPUTE=ON -DKOMPUTE_OPT_DISABLE_VULKAN_VERSION_CHECK=ON'
837829

838830
steps:
839831
- name: Clone
@@ -846,7 +838,7 @@ jobs:
846838
uses: hendrikmuhs/[email protected]
847839
with:
848840
key: windows-latest-cmake-${{ matrix.build }}
849-
variant: sccache
841+
variant: ccache
850842
evict-old-files: 1d
851843

852844
- name: Clone Kompute submodule
@@ -922,39 +914,26 @@ jobs:
922914
cp $env:RUNNER_TEMP/openblas/bin/libopenblas.dll ./build/bin/Release/openblas.dll
923915
cp $env:RUNNER_TEMP/OpenBLAS.LICENSE.txt ./build/bin/Release/OpenBLAS-${env:OPENBLAS_VERSION}.txt
924916
925-
- name: Check AVX512F support
926-
id: check_avx512f
927-
if: ${{ matrix.build == 'avx512-x64' }}
928-
continue-on-error: true
929-
run: |
930-
cd build
931-
$vcdir = $(vswhere -latest -products * -requires Microsoft.VisualStudio.Component.VC.Tools.x86.x64 -property installationPath)
932-
$msvc = $(join-path $vcdir $('VC\Tools\MSVC\'+$(gc -raw $(join-path $vcdir 'VC\Auxiliary\Build\Microsoft.VCToolsVersion.default.txt')).Trim()))
933-
$cl = $(join-path $msvc 'bin\Hostx64\x64\cl.exe')
934-
echo 'int main(void){unsigned int a[4];__cpuid(a,7);return !(a[1]&65536);}' >> avx512f.c
935-
& $cl /O2 /GS- /kernel avx512f.c /link /nodefaultlib /entry:main
936-
.\avx512f.exe && echo "AVX512F: YES" && ( echo HAS_AVX512F=1 >> $env:GITHUB_ENV ) || echo "AVX512F: NO"
937-
938917
- name: Test
939918
id: cmake_test
940-
# not all machines have native AVX-512
941-
if: ${{ matrix.build != 'msvc-arm64' && matrix.build != 'llvm-arm64' && matrix.build != 'llvm-arm64-opencl-adreno' && matrix.build != 'kompute-x64' && matrix.build != 'vulkan-x64' && (matrix.build != 'avx512-x64' || env.HAS_AVX512F == '1') }}
919+
if: ${{ matrix.build != 'llvm-arm64' && matrix.build != 'llvm-arm64-opencl-adreno' }}
942920
run: |
943921
cd build
944922
ctest -L main -C Release --verbose --timeout 900
945923
946-
- name: Test (Intel SDE)
947-
id: cmake_test_sde
948-
if: ${{ matrix.build == 'avx512-x64' && env.HAS_AVX512F == '0' }} # use Intel SDE for AVX-512 emulation
949-
run: |
950-
curl.exe -o $env:RUNNER_TEMP/sde.tar.xz -L "https://downloadmirror.intel.com/813591/sde-external-${env:SDE_VERSION}-win.tar.xz"
951-
# for some weird reason windows tar doesn't like sde tar.xz
952-
7z x "-o${env:RUNNER_TEMP}" $env:RUNNER_TEMP/sde.tar.xz
953-
7z x "-o${env:RUNNER_TEMP}" $env:RUNNER_TEMP/sde.tar
954-
$sde = $(join-path $env:RUNNER_TEMP sde-external-${env:SDE_VERSION}-win/sde.exe)
955-
cd build
956-
$env:LLAMA_SKIP_TESTS_SLOW_ON_EMULATOR = 1
957-
& $sde -future -- ctest -L main -C Release --verbose --timeout 900
924+
# TODO: disabled for now, consider adding tests for all CPU variants instead
925+
# - name: Test (Intel SDE)
926+
# id: cmake_test_sde
927+
# if: ${{ matrix.build == 'avx512-x64' && env.HAS_AVX512F == '0' }} # use Intel SDE for AVX-512 emulation
928+
# run: |
929+
# curl.exe -o $env:RUNNER_TEMP/sde.tar.xz -L "https://downloadmirror.intel.com/813591/sde-external-${env:SDE_VERSION}-win.tar.xz"
930+
# # for some weird reason windows tar doesn't like sde tar.xz
931+
# 7z x "-o${env:RUNNER_TEMP}" $env:RUNNER_TEMP/sde.tar.xz
932+
# 7z x "-o${env:RUNNER_TEMP}" $env:RUNNER_TEMP/sde.tar
933+
# $sde = $(join-path $env:RUNNER_TEMP sde-external-${env:SDE_VERSION}-win/sde.exe)
934+
# cd build
935+
# $env:LLAMA_SKIP_TESTS_SLOW_ON_EMULATOR = 1
936+
# & $sde -future -- ctest -L main -C Release --verbose --timeout 900
958937

959938
- name: Determine tag name
960939
id: tag
@@ -1039,7 +1018,7 @@ jobs:
10391018
uses: hendrikmuhs/[email protected]
10401019
with:
10411020
key: ${{ github.job }}-${{ matrix.cuda }}-${{ matrix.build }}
1042-
variant: sccache
1021+
variant: ccache
10431022
evict-old-files: 1d
10441023

10451024
- name: Install Cuda Toolkit 11.7
@@ -1117,6 +1096,8 @@ jobs:
11171096
cmake -S . -B build -G "Ninja Multi-Config" ^
11181097
-DLLAMA_BUILD_SERVER=ON ^
11191098
-DGGML_NATIVE=OFF ^
1099+
-DGGML_BACKEND_DL=ON ^
1100+
-DGGML_CPU_ALL_VARIANTS=ON ^
11201101
-DGGML_CUDA=ON ^
11211102
-DGGML_RPC=ON ^
11221103
-DCURL_LIBRARY="%CURL_PATH%/lib/libcurl.dll.a" -DCURL_INCLUDE_DIR="%CURL_PATH%/include"
@@ -1191,7 +1172,7 @@ jobs:
11911172
uses: hendrikmuhs/[email protected]
11921173
with:
11931174
key: windows-latest-cmake-sycl
1194-
variant: sccache
1175+
variant: ccache
11951176
evict-old-files: 1d
11961177

11971178
- name: Install

‎cmake/arm64-windows-msvc.cmake‎

Lines changed: 0 additions & 6 deletions
This file was deleted.

‎cmake/x64-windows-llvm.cmake‎

Lines changed: 0 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -3,9 +3,3 @@ set( CMAKE_SYSTEM_PROCESSOR x86_64 )
33

44
set( CMAKE_C_COMPILER clang )
55
set( CMAKE_CXX_COMPILER clang++ )
6-
7-
set( arch_c_flags "-march=native" )
8-
9-
set( CMAKE_C_FLAGS_INIT "${arch_c_flags}" )
10-
set( CMAKE_CXX_FLAGS_INIT "${arch_c_flags}" )
11-

‎tests/CMakeLists.txt‎

Lines changed: 4 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -114,7 +114,10 @@ if (NOT WIN32)
114114
target_include_directories(test-json-schema-to-grammar PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}/../tools/server)
115115
endif()
116116

117-
llama_build(test-quantize-stats.cpp)
117+
if (NOT GGML_BACKEND_DL)
118+
llama_build(test-quantize-stats.cpp)
119+
endif()
120+
118121
llama_build(test-gbnf-validator.cpp)
119122

120123
# build test-tokenizer-1-bpe target once and add many tests

0 commit comments

Comments
 (0)