Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

JIT TypeError: Object of type 'module' is not an instance of 'function' with Python 3.9.1

See original GitHub issue

🐛 Bug

Description

I cannot build the package on Arch Linux with PyTorch 1.7.1 and Python 3.9 (using aur , PKGBUILD)

Traceback (most recent call last):
  File "/home/trougnouf/.cache/pacaur/python-pytorch-piqa-git/src/piqa/setup.py", line 5, in <module>
    import piqa
  File "/home/trougnouf/.cache/pacaur/python-pytorch-piqa-git/src/piqa/piqa/__init__.py", line 10, in <module>
    from .tv import TV
  File "/home/trougnouf/.cache/pacaur/python-pytorch-piqa-git/src/piqa/piqa/tv.py", line 16, in <module>
    def tv(x: torch.Tensor, norm: str = 'L1') -> torch.Tensor:
  File "/usr/lib/python3.9/site-packages/torch/jit/_script.py", line 939, in script
    fn = torch._C._jit_script_compile(
TypeError: Object of type 'module' is not an instance of 'function'

This may be a bug in PyTorch JIT, incompatibility with newer Numpy, or other, but I’m reporting here since this is the only place I’ve encountered the bug so far and I haven’t seen another report.

Reproduce

On an up to date system:

git clone git@github.com:francois-rozet/piqa.git
cd piqa
python setup.py build

alternatively, with piqa succesfully built with older libraries and the system then updated, the same error is triggered with “import piqa”

Expected behavior

No error

Causes and solution

Workaround: set PIQA_JIT=0

Environment

PIQA version: 5994e349d4c140877e560b18c2a51307e61575fc
Python version: 3.9.1
OS: Arch Linux
Numpy 1.20
PyTorch 1.7.1

Issue Analytics

State:
Created 3 years ago
Comments:9 (5 by maintainers)

Top GitHub Comments

1reaction

trougnoufcommented, Mar 11, 2021

It’s working with pytorch 1.8.0 and python 3.9.2 😃

I guess it’ll be best to run with PIQA_JIT=1 since #7 .

Thank you!

0reactions

francois-rozetcommented, Mar 11, 2021

Hello @trougnouf 👋 There is a new version of torch (1.8.0) that was released. Maybe it solved the problem ? Nevertheless, I see that https://github.com/pytorch/pytorch/issues/51379 is still not resolved and the last comment corresponds exactly to your situation (Python 3.9.1, Arch, torch-1.7.1, pybind11-2.6.2, …). Since the problem isn’t caused by the PIQA package, I will close this issue. I will also merge #7, as I think it is preferable to not impose JITting by default.

Top Results From Across the Web

Module is not an instance of function - fastai

If I pip install fastai from github source on Debian/Bullseye using the --no-deps option (ensuring all requirements listed in settings.ini ...

What's New In Python 3.11 — Python 3.11.1 documentation

A TypeError is now raised instead of an AttributeError in with statements and contextlib.ExitStack.enter_context() for objects that do not support the ...

'Module' not callable with numba's jit compiler - Stack Overflow

A TypeError: 'module' object is not callable means you're trying to use something like a function when it's not actually a function or...

Release Notes — Numba 0.56.4+0.g288a38bbd.dirty-py3.7 ...

Version 0.56.2 (1 September, 2022) . This is a bugfix release that supports NumPy 1.23 and fixes CUDA function caching. Pull-Requests:.

Docs - CasADi

CasADi is available for C++, Python and MATLAB/Octave with little or no difference in ... SX.sym is a (static) function which returns an...