Can't install pytorch3d from source

[le-Greg]

🐛 Bugs / Unexpected behaviors

I tried to install pytorch3d in a docker container with pip, following this. Even though pytorch is installed i get this error :

ubuntu@9b041f0efe80:~$ python
Python 3.12.3 (main, Sep 11 2024, 14:17:37) [GCC 13.2.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import torch
>>> 
ubuntu@9b041f0efe80:~$ pip install "git+https://github.com/facebookresearch/pytorch3d.git"
Collecting git+https://github.com/facebookresearch/pytorch3d.git
  Cloning https://github.com/facebookresearch/pytorch3d.git to /tmp/pip-req-build-v2gs3l_2
  Running command git clone --filter=blob:none --quiet https://github.com/facebookresearch/pytorch3d.git /tmp/pip-req-build-v2gs3l_2
  Resolved https://github.com/facebookresearch/pytorch3d.git to commit 8ed0c7a00200df5ca08385b0f7c186b155175756
  Installing build dependencies ... done
  Getting requirements to build wheel ... error
  error: subprocess-exited-with-error
  
  × Getting requirements to build wheel did not run successfully.
  │ exit code: 1
  ╰─> [20 lines of output]
      Traceback (most recent call last):
        File "/home/ubuntu/venv/lib/python3.12/site-packages/pip/_vendor/pyproject_hooks/_in_process/_in_process.py", line 353, in <module>
          main()
        File "/home/ubuntu/venv/lib/python3.12/site-packages/pip/_vendor/pyproject_hooks/_in_process/_in_process.py", line 335, in main
          json_out['return_val'] = hook(**hook_input['kwargs'])
                                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
        File "/home/ubuntu/venv/lib/python3.12/site-packages/pip/_vendor/pyproject_hooks/_in_process/_in_process.py", line 118, in get_requires_for_build_wheel
          return hook(config_settings)
                 ^^^^^^^^^^^^^^^^^^^^^
        File "/tmp/pip-build-env-nepqkwa0/overlay/lib/python3.12/site-packages/setuptools/build_meta.py", line 332, in get_requires_for_build_wheel
          return self._get_build_requires(config_settings, requirements=[])
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
        File "/tmp/pip-build-env-nepqkwa0/overlay/lib/python3.12/site-packages/setuptools/build_meta.py", line 302, in _get_build_requires
          self.run_setup()
        File "/tmp/pip-build-env-nepqkwa0/overlay/lib/python3.12/site-packages/setuptools/build_meta.py", line 503, in run_setup
          super().run_setup(setup_script=setup_script)
        File "/tmp/pip-build-env-nepqkwa0/overlay/lib/python3.12/site-packages/setuptools/build_meta.py", line 318, in run_setup
          exec(code, locals())
        File "<string>", line 15, in <module>
      ModuleNotFoundError: No module named 'torch'
      [end of output]
  
  note: This error originates from a subprocess, and is likely not a problem with pip.
error: subprocess-exited-with-error

× Getting requirements to build wheel did not run successfully.
│ exit code: 1
╰─> See above for output.

note: This error originates from a subprocess, and is likely not a problem with pip.

Instructions To Reproduce the Issue:

Dockerfile:

FROM nvcr.io/nvidia/cuda:12.5.1-devel-ubuntu24.04

# Update apt and install necessary packages
ENV DEBIAN_FRONTEND=noninteractive
RUN apt-get update && apt-get install -y \
    build-essential libssl-dev zlib1g-dev libbz2-dev libreadline-dev \
    libsqlite3-dev wget curl llvm libncurses5-dev libncursesw5-dev \
    tk-dev libffi-dev liblzma-dev git software-properties-common \
    openssh-client sudo unzip apt-transport-https tmux

# Install Python
RUN add-apt-repository ppa:deadsnakes/ppa
RUN apt-get update && apt-get install -y python3.12 python3.12-venv python3-pip

# Switch to the non-root user
RUN echo "ubuntu ALL=(ALL) NOPASSWD:ALL" >> /etc/sudoers
USER ubuntu
WORKDIR /home/ubuntu

# Create and activate a virtual environment
RUN python3.12 -m venv /home/ubuntu/venv
ENV PATH="/home/ubuntu/venv/bin:$PATH"

# Upgrade pip in the virtual environment
RUN pip install --upgrade pip

# Install Python packages
RUN pip install torch torchvision
RUN pip install -U iopath

ENTRYPOINT ["bash"]

Then you can build and run with :

docker build -t pytorch3d .
docker run --gpus all -it pytorch3d

And then try to install pytorch3d inside the container :

curl -LO https://github.com/NVIDIA/cub/archive/1.10.0.tar.gz
tar xzf 1.10.0.tar.gz
export CUB_HOME=$PWD/cub-1.10.0
pip install "git+https://github.com/facebookresearch/pytorch3d.git"

[le-Greg]

I think it's because, when using "pip install ." with a setup.py file, and no pyproject.toml file, pip won't import the package that are installed. So no pytorch available.

You can clone locally and then use python setup.py install for short term, but i think using setup.py is going to be more and more deprecated. The other issue with this is that you can't list pytorch3d as a dependancy in your pyproject.toml, or i didn't find how.

[le-Greg]

Same issue as #1892
This works :

pip install --upgrade setuptools wheel
pip install "git+https://github.com/facebookresearch/pytorch3d.git@stable"