Skip to content

fix: Update Docker build to automatically adapt Torch version #1732

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 2 commits into from
Mar 28, 2023
Merged

fix: Update Docker build to automatically adapt Torch version #1732

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
ea6bdf4
fix: Refactor Docker Torch version to auto-sync
gs-olive Mar 10, 2023
11cf193
feat: Add build arg for CXX11 ABI
gs-olive Mar 28, 2023
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
4 changes: 1 addition & 3 deletions docker/Dockerfile
View file
Open in desktop
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Add build args for tensorrt and cudnn versions

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

use 22.04 base

Original file line number Diff line number Diff line change
Expand Up @@ -12,9 +12,6 @@ RUN ln -s /usr/bin/python3.8 /usr/bin/python
RUN python get-pip.py
RUN pip3 install wheel

# Install Pytorch
RUN pip3 install torch==2.0.0.dev20230103+cu117 torchvision==0.15.0.dev20230103+cu117 --extra-index-url https://download.pytorch.org/whl/nightly/cu117

# Install CUDNN + TensorRT
RUN wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/cuda-ubuntu2004.pin
RUN mv cuda-ubuntu2004.pin /etc/apt/preferences.d/cuda-repository-pin-600
Expand Down Expand Up @@ -64,6 +61,7 @@ COPY . /opt/torch_tensorrt
COPY --from=torch-tensorrt-builder /workspace/torch_tensorrt/src/py/dist/ .

RUN cp /opt/torch_tensorrt/docker/WORKSPACE.docker /opt/torch_tensorrt/WORKSPACE
RUN pip install -r /opt/torch_tensorrt/py/requirements.txt
RUN pip3 install *.whl && rm -fr /workspace/torch_tensorrt/py/dist/* *.whl

# Install native tensorrt python package required by torch_tensorrt whl file
Expand Down
5 changes: 3 additions & 2 deletions docker/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,7 +5,8 @@
* `Dockerfile` currently uses the exact library versions (Torch, CUDA, CUDNN, TensorRT) listed in <a href="https://github.com/pytorch/TensorRT#dependencies">dependencies</a> to build Torch-TensorRT.

* This `Dockerfile` installs `pre-cxx11-abi` versions of Pytorch and builds Torch-TRT using `pre-cxx11-abi` libtorch as well.
Note: To install `cxx11_abi` version of Torch-TensorRT, enable `USE_CXX11=1` flag so that `dist-build.sh` can build it accordingly.

Note: Torch-TRT based on the C++11 ABI is not automatically installed in the build. To install it, modify the `WORKSPACE.docker` file to specify a C++11 ABI path, then enable the `USE_CXX11=1` flag so that `dist-build.sh` can build it accordingly. See the `WORKSPACE` file in the root of the repository for an example.

### Dependencies

Expand Down Expand Up @@ -38,4 +39,4 @@ bazel test //tests/core/conversion/converters:test_activation --compilation_mode

### Pytorch NGC containers

We also ship Torch-TensorRT in <a href="https://catalog.ngc.nvidia.com/orgs/nvidia/containers/pytorch">Pytorch NGC containers </a>. Release notes for these containers can be found <a href="https://docs.nvidia.com/deeplearning/frameworks/pytorch-release-notes/index.html">here</a>. Check out `release/ngc/23.XX` branch of Torch-TensorRT for source code that gets shipped with `23.XX` version of Pytorch NGC container.
We also ship Torch-TensorRT in <a href="https://catalog.ngc.nvidia.com/orgs/nvidia/containers/pytorch">Pytorch NGC containers </a>. Release notes for these containers can be found <a href="https://docs.nvidia.com/deeplearning/frameworks/pytorch-release-notes/index.html">here</a>. Check out `release/ngc/23.XX` branch of Torch-TensorRT for source code that gets shipped with `23.XX` version of Pytorch NGC container.
16 changes: 6 additions & 10 deletions docker/WORKSPACE.docker
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -48,20 +48,16 @@ new_local_repository(
# Tarballs and fetched dependencies (default - use in cases when building from precompiled bin and tarballs)
#############################################################################################################

http_archive(
new_local_repository(
name = "libtorch",
build_file = "@//third_party/libtorch:BUILD",
sha256 = "59b8b5e1954a86d50b79c13f06398d385b200da13e37a08ecf31d3c62e5ca127",
strip_prefix = "libtorch",
urls = ["https://download.pytorch.org/libtorch/nightly/cu117/libtorch-cxx11-abi-shared-with-deps-2.0.0.dev20230103%2Bcu117.zip"],
path = "/usr/local/lib/python3.8/dist-packages/torch/",
build_file = "third_party/libtorch/BUILD"
)

http_archive(
new_local_repository(
name = "libtorch_pre_cxx11_abi",
build_file = "@//third_party/libtorch:BUILD",
sha256 = "e260fc7476be89d1650953e8643e9f7363845f5a52de4bab87ac0e619c1f6ad4",
strip_prefix = "libtorch",
urls = ["https://download.pytorch.org/libtorch/nightly/cu117/libtorch-shared-with-deps-2.0.0.dev20230103%2Bcu117.zip"],
path = "/usr/local/lib/python3.8/dist-packages/torch/",
build_file = "third_party/libtorch/BUILD"
)

####################################################################################
Expand Down