Broken wiki, no place to download?

[colingoodman]

Pip3 failed to install pycuda, so I found the linked wiki to troubleshoot. It explains I need to download a tgz pycuda file, but all of the links appear to be broken:

This is the case on both the linux and mac pages of the wiki.

[inducer]

I agree that the Wiki situation is a bit sad. What I did there was a minimal job to salvage some wiki pages from a moinmoin wiki that was on the way out. That documentation should be moved to the docs.

As for download, just like pip, download from the package index:

https://pypi.org/project/pycuda/

And if you run into installation trouble, feel free to post here (with detailed error messages).

[colingoodman]

Thanks for the help. When attempting to install via pip, I get ERROR: Could not build wheels for pycuda, which is required to install pyproject.toml-based projects

I've installed wheel via pip as well as the following apt packages: software-properties-common libboost-all-dev python-numpy python-setuptools libboost-python-dev libboost-thread-dev python3-dev

I am attempting to install pycuda into a docker container via dockerfile.

[colingoodman]

Attempting to install pycuda onto a Nvidia jetson device on jetpack 5, I receive some more verbose feedback from pip.

Some example stuff:

      src/cpp/cuda.hpp:14:10: fatal error: cuda.h: No such file or directory
         14 | #include <cuda.h>
            |          ^~~~~~~~
      compilation terminated.
      /tmp/pip-build-env-hn6xrr2x/overlay/lib/python3.8/site-packages/setuptools/command/build_py.py:202: SetuptoolsDeprecationWarning:     Installing 'pycuda.cuda' as data is deprecated, please list it in `packages`.

Python recognizes 'pycuda.cuda' as an importable package,
          but it is not listed in the `packages` configuration of setuptools.

ERROR: Could not build wheels for pycuda, which is required to install pyproject.toml-based projects

[mitkotak]

Can you try ?

pip install git+https://github.com/inducer/pycuda.git@main

[colingoodman]

@mitkotak Thanks for the suggestion, unfortunately it did not change anything.

I did find a forum thread that suggests setting some environment variables. This fixed the installation on my Nvidia device, but not on my Docker build. This implies to me that there is a lack of environment parity with the docker build and I thus need to configure/install something else.