matt/ollama37
1
0
Fork 0
mirror of https://github.com/dogkeeper886/ollama37.git synced 2025-12-12 00:37:04 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'ollama:main' into main

Browse Source
This commit is contained in:
mraiser
2024-01-27 21:56:11 -05:00
committed by GitHub
parent 6eb3cddcb6 e02ecfb6c8
commit 4c4c730a0a
25 changed files with 739 additions and 195 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
llm/generate/gen_common.sh
View File

@@ -39,6 +39,9 @@ init_vars() {
*)
;;
esac
if [ -z "${CMAKE_CUDA_ARCHITECTURES}" ] ; then
CMAKE_CUDA_ARCHITECTURES="50;52;61;70;75;80"
fi
}
git_module_setup() {
@@ -61,6 +64,17 @@ apply_patches() {
if ! grep ollama ${LLAMACPP_DIR}/examples/server/CMakeLists.txt; then
echo 'include (../../../ext_server/CMakeLists.txt) # ollama' >>${LLAMACPP_DIR}/examples/server/CMakeLists.txt
fi
# apply temporary patches until fix is upstream
for patch in ../patches/*.diff; do
for file in $(grep "^+++ " ${patch} | cut -f2 -d' ' | cut -f2- -d/); do
(cd ${LLAMACPP_DIR}; git checkout ${file})
done
done
for patch in ../patches/*.diff; do
(cd ${LLAMACPP_DIR} && git apply ${patch})
done
# Avoid duplicate main symbols when we link into the cgo binary
sed -e 's/int main(/int __main(/g' <${LLAMACPP_DIR}/examples/server/server.cpp >${LLAMACPP_DIR}/examples/server/server.cpp.tmp &&
mv ${LLAMACPP_DIR}/examples/server/server.cpp.tmp ${LLAMACPP_DIR}/examples/server/server.cpp

2
llm/generate/gen_linux.sh
View File

@@ -140,7 +140,7 @@ if [ -d "${CUDA_LIB_DIR}" ]; then
if [ -n "${CUDA_MAJOR}" ]; then
CUDA_VARIANT=_v${CUDA_MAJOR}
fi
CMAKE_DEFS="-DLLAMA_CUBLAS=on ${COMMON_CMAKE_DEFS} ${CMAKE_DEFS}"
CMAKE_DEFS="-DLLAMA_CUBLAS=on -DLLAMA_CUDA_FORCE_MMQ=on -DCMAKE_CUDA_ARCHITECTURES=${CMAKE_CUDA_ARCHITECTURES} ${COMMON_CMAKE_DEFS} ${CMAKE_DEFS}"
BUILD_DIR="${LLAMACPP_DIR}/build/linux/${ARCH}/cuda${CUDA_VARIANT}"
EXTRA_LIBS="-L${CUDA_LIB_DIR} -lcudart -lcublas -lcublasLt -lcuda"
build

30
llm/generate/gen_windows.ps1
View File

@@ -25,6 +25,11 @@ function init_vars {
}
$script:GZIP=(get-command -ea 'silentlycontinue' gzip).path
$script:DUMPBIN=(get-command -ea 'silentlycontinue' dumpbin).path
if ($null -eq $env:CMAKE_CUDA_ARCHITECTURES) {
$script:CMAKE_CUDA_ARCHITECTURES="50;52;61;70;75;80"
} else {
$script:CMAKE_CUDA_ARCHITECTURES=$env:CMAKE_CUDA_ARCHITECTURES
}
}
function git_module_setup {
@@ -40,6 +45,29 @@ function apply_patches {
if (!(Select-String -Path "${script:llamacppDir}/examples/server/CMakeLists.txt" -Pattern 'ollama')) {
Add-Content -Path "${script:llamacppDir}/examples/server/CMakeLists.txt" -Value 'include (../../../ext_server/CMakeLists.txt) # ollama'
}
# Apply temporary patches until fix is upstream
$patches = Get-ChildItem "../patches/*.diff"
foreach ($patch in $patches) {
# Extract file paths from the patch file
$filePaths = Get-Content $patch.FullName | Where-Object { $_ -match '^\+\+\+ ' } | ForEach-Object {
$parts = $_ -split ' '
($parts[1] -split '/', 2)[1]
}
# Checkout each file
foreach ($file in $filePaths) {
Set-Location -Path ${script:llamacppDir}
git checkout $file
}
}
# Apply each patch
foreach ($patch in $patches) {
Set-Location -Path ${script:llamacppDir}
git apply $patch.FullName
}
# Avoid duplicate main symbols when we link into the cgo binary
$content = Get-Content -Path "${script:llamacppDir}/examples/server/server.cpp"
$content = $content -replace 'int main\(', 'int __main('
@@ -128,7 +156,7 @@ if ($null -ne $script:CUDA_LIB_DIR) {
}
init_vars
$script:buildDir="${script:llamacppDir}/build/windows/${script:ARCH}/cuda$script:CUDA_VARIANT"
$script:cmakeDefs += @("-DLLAMA_CUBLAS=ON", "-DLLAMA_AVX=on")
$script:cmakeDefs += @("-DLLAMA_CUBLAS=ON", "-DLLAMA_AVX=on", "-DCMAKE_CUDA_ARCHITECTURES=${script:CMAKE_CUDA_ARCHITECTURES}")
build
install
cp "${script:CUDA_LIB_DIR}/cudart64_*.dll" "${script:buildDir}/lib"