matt/ollama37
1
0
Fork 0
mirror of https://github.com/dogkeeper886/ollama37.git synced 2025-12-11 00:07:07 +00:00
Code Issues Packages Projects Releases Wiki Activity

Split rocm back out of bundle (#6432)

Browse Source 
We're over budget for github's maximum release artifact size with rocm + 2 cuda
versions.  This splits rocm back out as a discrete artifact, but keeps the layout so it can
be extracted into the same location as the main bundle.
This commit is contained in:
Daniel Hiltgen
2024-08-20 07:26:38 -07:00
committed by GitHub
parent 19e5a890f7
commit a017cf2fea
5 changed files with 16 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
llm/generate/gen_linux.sh
View File

@@ -260,7 +260,8 @@ if [ -z "${OLLAMA_SKIP_ROCM_GENERATE}" -a -d "${ROCM_PATH}" ]; then
echo "Building custom ROCM GPU"
fi
BUILD_DIR="../build/linux/${ARCH}/rocm${ROCM_VARIANT}"
ROCM_DIST_DIR="${DIST_BASE}/lib/ollama"
# ROCm dependencies are too large to fit into a unified bundle
ROCM_DIST_DIR="${DIST_BASE}/../linux-${GOARCH}-rocm/lib/ollama"
# TODO figure out how to disable runpath (rpath)
# export CMAKE_HIP_FLAGS="-fno-rtlib-add-rpath" # doesn't work
export LLAMA_SERVER_LDFLAGS="-L${ROCM_PATH}/lib -L/opt/amdgpu/lib/x86_64-linux-gnu/ -lhipblas -lrocblas -lamdhip64 -lrocsolver -lamd_comgr -lhsa-runtime64 -lrocsparse -ldrm -ldrm_amdgpu"