Torch is not found in Dockerfile

[Tickets14]

Error when building a detectron2 on docker
Hello. Im experiencing a weird behavior where if im building a docker with detectron2 it always showing an error

3.930   × python setup.py egg_info did not run successfully.
3.930   │ exit code: 1
3.930   ╰─> [6 lines of output]
3.930       Traceback (most recent call last):
3.930         File "<string>", line 2, in <module>
3.930         File "<pip-setuptools-caller>", line 34, in <module>
3.930         File "/tmp/pip-req-build-sz8hbgh3/setup.py", line 10, in <module>
3.930           import torch
3.930       ModuleNotFoundError: No module named 'torch'
3.930       [end of output]

Here is my Dokerfile

FROM public.ecr.aws/lambda/python:3.9


# Install Git
RUN yum install -y git

# Copy requirements.txt
COPY requirements.txt ./

# Install the specified packages
RUN pip3 install -r requirements.txt;


COPY sample.py ./
CMD [ "sampl.handler"]

Here is my requirements.txt

boto3==1.28.62
Pillow==10.0.1
numpy==1.24.3
torch==2.0.0
torchvision==0.15.1
opencv-python==4.6.0.66
git+https://github.com/facebookresearch/detectron2.git

[github-actions]

You've chosen to report an unexpected problem or bug. Unless you already know the root cause of it, please include details about it by filling the issue template.
The following information is missing: "Instructions To Reproduce the Issue and Full Logs";

[Tickets14]

I also tried this on my docker

FROM public.ecr.aws/lambda/python:3.9


# Install Git
RUN yum install -y git

# Copy requirements.txt
COPY requirements.txt ./

# Install the specified packages
RUN pip3 install -r requirements.txt;
RUN git clone https://github.com/facebookresearch/detectron2.git
RUN pip3 install -e detectron2

COPY sample.py ./
CMD [ "sample.handler"]

Here is the error

21.34   Running setup.py install for antlr4-python3-runtime: started
21.49   Running setup.py install for antlr4-python3-runtime: finished with status 'done'
21.83   Attempting uninstall: iopath
21.83     Found existing installation: iopath 0.1.10
21.83     Uninstalling iopath-0.1.10:
21.94       Successfully uninstalled iopath-0.1.10
22.75   Attempting uninstall: fvcore
22.75     Found existing installation: fvcore 0.1.6
22.75     Uninstalling fvcore-0.1.6:
22.82       Successfully uninstalled fvcore-0.1.6
22.82   DEPRECATION: fvcore is being installed using the legacy 'setup.py install' method, because it does not have a 'pyproject.toml' and the 'wheel' package is not installed. pip 23.1 will enforce this behaviour change. A possible replacement is to enable the '--use-pep517' option. Discussion can be found at https://github.com/pypa/pip/issues/8559
22.82   Running setup.py install for fvcore: started
22.96   Running setup.py install for fvcore: finished with status 'done'
24.18   Running setup.py develop for detectron2
25.38     error: subprocess-exited-with-error
25.38     
25.38     × python setup.py develop did not run successfully.
25.38     │ exit code: 1
25.38     ╰─> [15 lines of output]
25.38         running develop
25.38         running egg_info
25.38         creating detectron2.egg-info
25.38         writing detectron2.egg-info/PKG-INFO
25.38         writing dependency_links to detectron2.egg-info/dependency_links.txt
25.38         writing requirements to detectron2.egg-info/requires.txt
25.38         writing top-level names to detectron2.egg-info/top_level.txt
25.38         writing manifest file 'detectron2.egg-info/SOURCES.txt'
25.38         /var/lang/lib/python3.9/site-packages/torch/utils/cpp_extension.py:476: UserWarning: Attempted to use ninja as the BuildExtension backend but we could not find ninja.. Falling back to using the slow distutils backend.
25.38           warnings.warn(msg.format('we could not find ninja.'))
25.38         reading manifest file 'detectron2.egg-info/SOURCES.txt'
25.38         adding license file 'LICENSE'
25.38         writing manifest file 'detectron2.egg-info/SOURCES.txt'
25.38         running build_ext
25.38         error: [Errno 2] No such file or directory: 'which'
25.38         [end of output]
25.38     
25.38     note: This error originates from a subprocess, and is likely not a problem with pip.
25.38 error: subprocess-exited-with-error
25.38 
25.38 × python setup.py develop did not run successfully.
25.38 │ exit code: 1
25.38 ╰─> [15 lines of output]
25.38     running develop
25.38     running egg_info
25.38     creating detectron2.egg-info
25.38     writing detectron2.egg-info/PKG-INFO
25.38     writing dependency_links to detectron2.egg-info/dependency_links.txt
25.38     writing requirements to detectron2.egg-info/requires.txt
25.38     writing top-level names to detectron2.egg-info/top_level.txt
25.38     writing manifest file 'detectron2.egg-info/SOURCES.txt'
25.38     /var/lang/lib/python3.9/site-packages/torch/utils/cpp_extension.py:476: UserWarning: Attempted to use ninja as the BuildExtension backend but we could not find ninja.. Falling back to using the slow distutils backend.
25.38       warnings.warn(msg.format('we could not find ninja.'))
25.38     reading manifest file 'detectron2.egg-info/SOURCES.txt'
25.38     adding license file 'LICENSE'
25.38     writing manifest file 'detectron2.egg-info/SOURCES.txt'
25.38     running build_ext
25.38     error: [Errno 2] No such file or directory: 'which'
25.38     [end of output]
25.38 
25.38 note: This error originates from a subprocess, and is likely not a problem with pip.```

[Prayag22]

FROM` public.ecr.aws/lambda/python:3.9
# Install packages
RUN pip install torch==2.0.0

# Install Git
RUN yum install -y git

# Copy requirements.txt
COPY requirements.txt ./

# Install the specified packages
RUN pip3 install -r requirements.txt;
RUN git clone https://github.com/facebookresearch/detectron2.git
RUN pip3 install -e detectron2

COPY sample.py ./
CMD [ "sample.handler"]```

[brownaaron45]

I wanted to post a solution that I found for this problem. It is not a problem with detection but rather a problem with your torch library. Since you are using lambda which is CPU only you need to specifically select CPU wheel files for pytorch.

https://pytorch.org/get-started/locally/#start-locally

RUN pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu \
    && python -c "import torch; print(torch.__version__)"