Skip to content

setting MAX_JOBS=5 for oom in CUDA arm wheel #1828

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
May 22, 2024
Merged

setting MAX_JOBS=5 for oom in CUDA arm wheel #1828

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 5 additions & 5 deletions aarch64_linux/aarch64_ci_build.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -28,10 +28,10 @@ cd /
git config --global --add safe.directory /pytorch
pip install -r /pytorch/requirements.txt
pip install auditwheel
if [ -n "$GPU_ARCH_VERSION" ]; then
echo "BASE_CUDA_VERSION is set to: $GPU_ARCH_VERSION"
python /builder/aarch64_linux/aarch64_wheel_ci_build.py --enable-mkldnn --enable-cuda
else
echo "BASE_CUDA_VERSION is not set."
if [ "$DESIRED_CUDA" = "cpu" ]; then
echo "BASE_CUDA_VERSION is not set. Building cpu wheel."
python /builder/aarch64_linux/aarch64_wheel_ci_build.py --enable-mkldnn
else
echo "BASE_CUDA_VERSION is set to: $DESIRED_CUDA"
python /builder/aarch64_linux/aarch64_wheel_ci_build.py --enable-mkldnn --enable-cuda
fi
13 changes: 6 additions & 7 deletions aarch64_linux/aarch64_wheel_ci_build.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -122,12 +122,10 @@ def update_wheel(wheel_path) -> None:
"/usr/local/cuda/lib64/libcudnn_cnn_train.so.8",
"/usr/local/cuda/lib64/libcudnn_ops_infer.so.8",
"/usr/local/cuda/lib64/libcudnn_ops_train.so.8",
"/opt/conda/envs/aarch64_env/lib/libopenblas.so.0",
"/opt/conda/envs/aarch64_env/lib/libgfortran.so.5",
"/opt/conda/envs/aarch64_env/lib/libgomp.so.1",
"/opt/OpenBLAS/lib/libopenblas.so.0",
"/acl/build/libarm_compute.so",
"/acl/build/libarm_compute_graph.so",
"/acl/build/libarm_compute_core.so",
]
# Copy libraries to unzipped_folder/a/lib
for lib_path in libs_to_copy:
Expand All @@ -140,10 +138,10 @@ def update_wheel(wheel_path) -> None:
os.system(f"cd {folder}/tmp/; zip -r {folder}/cuda_wheel/{wheelname} *")
shutil.move(
f"{folder}/cuda_wheel/{wheelname}",
f"/dist/{wheelname}",
f"{folder}/{wheelname}",
copy_function=shutil.copy2,
)
os.system(f"rm -rf {folder}/tmp {folder}/dist/cuda_wheel/")
os.system(f"rm -rf {folder}/tmp/ {folder}/cuda_wheel/")


def complete_wheel(folder: str) -> str:
Expand Down Expand Up @@ -201,8 +199,9 @@ def parse_arguments():
branch = "master"

print("Building PyTorch wheel")
build_vars = "CMAKE_SHARED_LINKER_FLAGS=-Wl,-z,max-page-size=0x10000 "
os.system("python setup.py clean")
os.system("export USE_PRIORITIZED_TEXT_FOR_LD=1") #enable linker script optimization https://github.com/pytorch/pytorch/pull/121975/files
build_vars = "MAX_JOBS=5 CMAKE_SHARED_LINKER_FLAGS=-Wl,-z,max-page-size=0x10000 "
os.system("cd /pytorch; python setup.py clean")

override_package_version = os.getenv("OVERRIDE_PACKAGE_VERSION")
if override_package_version is not None:
Expand Down
6 changes: 5 additions & 1 deletion manywheel/build_cuda.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -60,7 +60,11 @@ cuda_version_nodot=$(echo $CUDA_VERSION | tr -d '.')
TORCH_CUDA_ARCH_LIST="5.0;6.0;7.0;7.5;8.0;8.6"
case ${CUDA_VERSION} in
12.4)
TORCH_CUDA_ARCH_LIST="${TORCH_CUDA_ARCH_LIST};9.0"
if [[ "$GPU_ARCH_TYPE" = "cuda-aarch64" ]]; then
TORCH_CUDA_ARCH_LIST="9.0"
else
TORCH_CUDA_ARCH_LIST="${TORCH_CUDA_ARCH_LIST};9.0"
fi
EXTRA_CAFFE2_CMAKE_FLAGS+=("-DATEN_NO_TEST=ON")
;;
12.1)
Expand Down
Loading