Skip to content

[ML] Fix PyTorch Docker builds: copy libtorch from installed package #3091

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
edsavage merged 2 commits into from
Jul 23, 2026
+26 −11
Merged

[ML] Fix PyTorch Docker builds: copy libtorch from installed package #3091

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
8bdeabe
[ML] Fix PyTorch Docker builds: copy libtorch from installed package
edsavage Jul 22, 2026
394d944
[ML] Use mkdir -p for pytorch include dir in pinned images
edsavage Jul 22, 2026
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
13 changes: 9 additions & 4 deletions dev-tools/docker/linux_aarch64_native_image/Dockerfile
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -187,10 +187,15 @@ RUN \
export PYTORCH_BUILD_NUMBER=1 && \
export MAX_JOBS=10 && \
/usr/local/bin/python3.12 setup.py install && \
mkdir /usr/local/gcc133/include/pytorch && \
cp -r torch/include/* /usr/local/gcc133/include/pytorch/ && \
cp torch/lib/libtorch_cpu.so /usr/local/gcc133/lib && \
cp torch/lib/libc10.so /usr/local/gcc133/lib && \
# Recent PyTorch redirects `setup.py install` to `pip install .`, which installs the
# torch package (headers + libs) into site-packages rather than leaving them in the
# source tree (see pytorch/pytorch#152276). Resolve the installed package location so
# this works whether the build is in-tree (older tags) or a wheel install (newer).
TORCH_DIR=$(/usr/local/bin/python3.12 -c "import sysconfig, os; print(os.path.join(sysconfig.get_paths()['platlib'], 'torch'))") && \
mkdir -p /usr/local/gcc133/include/pytorch && \
cp -r "${TORCH_DIR}"/include/* /usr/local/gcc133/include/pytorch/ && \
cp "${TORCH_DIR}"/lib/libtorch_cpu.so /usr/local/gcc133/lib && \
cp "${TORCH_DIR}"/lib/libc10.so /usr/local/gcc133/lib && \
Comment thread
edsavage marked this conversation as resolved.
cd .. && \
rm -rf pytorch /tmp/*

Expand Down
13 changes: 9 additions & 4 deletions dev-tools/docker/linux_image/Dockerfile
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -202,10 +202,15 @@ RUN \
export PYTORCH_BUILD_NUMBER=1 && \
export MAX_JOBS=10 && \
/usr/local/bin/python3.12 setup.py install && \
mkdir /usr/local/gcc133/include/pytorch && \
cp -r torch/include/* /usr/local/gcc133/include/pytorch/ && \
cp torch/lib/libtorch_cpu.so /usr/local/gcc133/lib && \
cp torch/lib/libc10.so /usr/local/gcc133/lib && \
# Recent PyTorch redirects `setup.py install` to `pip install .`, which installs the
# torch package (headers + libs) into site-packages rather than leaving them in the
# source tree (see pytorch/pytorch#152276). Resolve the installed package location so
# this works whether the build is in-tree (older tags) or a wheel install (newer).
TORCH_DIR=$(/usr/local/bin/python3.12 -c "import sysconfig, os; print(os.path.join(sysconfig.get_paths()['platlib'], 'torch'))") && \
mkdir -p /usr/local/gcc133/include/pytorch && \
cp -r "${TORCH_DIR}"/include/* /usr/local/gcc133/include/pytorch/ && \
cp "${TORCH_DIR}"/lib/libtorch_cpu.so /usr/local/gcc133/lib && \
cp "${TORCH_DIR}"/lib/libc10.so /usr/local/gcc133/lib && \
Comment thread
edsavage marked this conversation as resolved.
cd .. && \
rm -rf pytorch /tmp/*

Expand Down
11 changes: 8 additions & 3 deletions dev-tools/docker/pytorch_linux_image/Dockerfile
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -108,10 +108,15 @@ RUN --mount=type=secret,id=gcs_key \
/usr/local/bin/python3.12 -m pip install --trusted-host pypi.python.org --trusted-host files.pythonhosted.org --trusted-host pypi.org -r requirements.txt && \
/usr/local/bin/python3.12 setup.py install && \
if command -v sccache &>/dev/null && sccache --show-stats &>/dev/null; then sccache --show-stats; sccache --stop-server; fi && \
# Recent PyTorch redirects `setup.py install` to `pip install .`, which installs the
# torch package (headers + libs) into site-packages rather than leaving them in the
# source tree (see pytorch/pytorch#152276). Resolve the installed package location so
# this works whether the build is in-tree (older tags) or a wheel install (newer).
TORCH_DIR=$(/usr/local/bin/python3.12 -c "import sysconfig, os; print(os.path.join(sysconfig.get_paths()['platlib'], 'torch'))") && \
mkdir -p /usr/local/gcc133/include/pytorch && \
/bin/cp -rf torch/include/* /usr/local/gcc133/include/pytorch/ && \
/bin/cp -f torch/lib/libtorch_cpu.so /usr/local/gcc133/lib && \
/bin/cp -f torch/lib/libc10.so /usr/local/gcc133/lib && \
/bin/cp -rf "${TORCH_DIR}"/include/* /usr/local/gcc133/include/pytorch/ && \
/bin/cp -f "${TORCH_DIR}"/lib/libtorch_cpu.so /usr/local/gcc133/lib && \
/bin/cp -f "${TORCH_DIR}"/lib/libc10.so /usr/local/gcc133/lib && \
cd .. && \
rm -rf pytorch

Expand Down
Loading