Import error (numpy.ufunc) with python 2.7 and numpy 1.16.2

[edrozenberg]

Description

Trying to import skimage causes AttributeError: 'numpy.ufunc' object has no attribute '__module__' when using python 2.7, skimage 0.14.2, numpy 1.16.2 .

Way to reproduce

import skimage

Version information

# Paste the output of the following python commands
from __future__ import print_function
import sys; print(sys.version)
import platform; print(platform.platform())
import skimage; print("scikit-image version: {}".format(skimage.__version__))
import numpy; print("numpy version: {}".format(numpy.__version__))

2.7.16 (default, Mar  3 2019, 19:53:40) 
[GCC 8.3.0]
Linux-4.19.30-x86_64-Intel-R-_Xeon-R-_CPU_E5-2630_v3_@_2.40GHz-with-slackware-14.2

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/lib64/python2.7/site-packages/skimage/__init__.py", line 167, in <module>
    from .util.dtype import (img_as_float32,
  File "/usr/lib64/python2.7/site-packages/skimage/util/__init__.py", line 6, in <module>
    from .apply_parallel import apply_parallel
  File "/usr/lib64/python2.7/site-packages/skimage/util/apply_parallel.py", line 8, in <module>
    import dask.array as da
  File "/usr/lib64/python2.7/site-packages/dask/array/__init__.py", line 9, in <module>
    from .routines import (take, choose, argwhere, where, coarsen, insert,
  File "/usr/lib64/python2.7/site-packages/dask/array/routines.py", line 256, in <module>
    @wraps(np.matmul)
  File "/usr/lib64/python2.7/functools.py", line 33, in update_wrapper
    setattr(wrapper, attr, getattr(wrapped, attr))
AttributeError: 'numpy.ufunc' object has no attribute '__module__'

numpy version: 1.16.2

• Problem only happens with python 2.7. Works fine under python 3.7. Same versions of everything.
• numpy was compiled from source for both python2 and python3
• scikit-image - same issue whether I use precompiled from pip2 or compile from source
• I double-checked that I don't have multiple versions of anything
• The only thing that fixes things under python2 so far is pip2 install scikit-image==0.12.2

# pip2 list|grep numpy
numpy 1.16.2          

# pip3 list|grep numpy
numpy                1.16.2   

[stefanv]

Looks like you need to upgrade on 2.7 to.0.14.2.

[edrozenberg]

Thanks but 0.14.2 is what causes the problem. I was just showing that 0.12.2 is what works, and that is what I have installed now after 0.14.2 failed to work. I'll edit my post to remove that confusing info.

[edrozenberg]

I'll have to try 0.15dev and see if that makes any difference.

[stefanv]

Sorry, I didn't check properly: the error you are seeing is coming from dask.

[stefanv]

(I.e., you should first try to upgrade dask.)

[edrozenberg]

Brilliant! Was not aware dask was involved. Fixed by updating dask:

# pip2 install -U dask

Installing collected packages: dask
  Found existing installation: dask 1.0.0
    Uninstalling dask-1.0.0:
      Successfully uninstalled dask-1.0.0
Successfully installed dask-1.1.4

[ranka47]

sudo pip install -U dask numpy scikit-image
Requirement already up-to-date: dask in /usr/local/lib/python2.7/dist-packages (1.2.2)
Requirement already up-to-date: numpy in /usr/local/lib/python2.7/dist-packages (1.16.4)
Requirement already up-to-date: scikit-image in /usr/local/lib/python2.7/dist-packages (0.14.3)
Requirement already satisfied, skipping upgrade: cloudpickle>=0.2.1 in /usr/local/lib/python2.7/dist-packages (from scikit-image) (1.2.1)
Requirement already satisfied, skipping upgrade: pillow>=4.3.0 in /usr/local/lib/python2.7/dist-packages (from scikit-image) (6.1.0)
Requirement already satisfied, skipping upgrade: scipy>=0.17.0 in /usr/lib/python2.7/dist-packages (from scikit-image) (0.17.0)
Requirement already satisfied, skipping upgrade: six>=1.10.0 in /usr/lib/python2.7/dist-packages (from scikit-image) (1.10.0)
Requirement already satisfied, skipping upgrade: PyWavelets>=0.4.0 in /usr/local/lib/python2.7/dist-packages (from scikit-image) (1.0.3)
Requirement already satisfied, skipping upgrade: networkx>=1.8 in /usr/local/lib/python2.7/dist-packages (from scikit-image) (2.2)
Requirement already satisfied, skipping upgrade: matplotlib>=2.0.0 in /usr/local/lib/python2.7/dist-packages (from scikit-image) (2.2.4)
Requirement already satisfied, skipping upgrade: decorator>=4.3.0 in /usr/local/lib/python2.7/dist-packages (from networkx>=1.8->scikit-image) (4.4.0)
Requirement already satisfied, skipping upgrade: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in /usr/lib/python2.7/dist-packages (from matplotlib>=2.0.0->scikit-image) (2.0.3)
Requirement already satisfied, skipping upgrade: backports.functools-lru-cache in /usr/local/lib/python2.7/dist-packages (from matplotlib>=2.0.0->scikit-image) (1.5)
Requirement already satisfied, skipping upgrade: subprocess32 in /usr/local/lib/python2.7/dist-packages (from matplotlib>=2.0.0->scikit-image) (3.5.4)
Requirement already satisfied, skipping upgrade: pytz in /usr/lib/python2.7/dist-packages (from matplotlib>=2.0.0->scikit-image) (2014.10)
Requirement already satisfied, skipping upgrade: python-dateutil>=2.1 in /usr/lib/python2.7/dist-packages (from matplotlib>=2.0.0->scikit-image) (2.4.2)
Requirement already satisfied, skipping upgrade: kiwisolver>=1.0.1 in /usr/local/lib/python2.7/dist-packages (from matplotlib>=2.0.0->scikit-image) (1.1.0)
Requirement already satisfied, skipping upgrade: cycler>=0.10 in /usr/local/lib/python2.7/dist-packages (from matplotlib>=2.0.0->scikit-image) (0.10.0)
Requirement already satisfied, skipping upgrade: setuptools in /usr/lib/python2.7/dist-packages (from kiwisolver>=1.0.1->matplotlib>=2.0.0->scikit-image) (20.7.0)

Python version: 2.7.15
numpy Version: 1.16.4
dask Version: 1.2.2
scikit-image Version: 0.14.3

I am getting the same error.

Is there any other solution to this? I upgraded all three packages.
It worked before with
numpy version: 1.12.0
scikit-image: 0.14.1
dask: 1.0.0
However, when I upgraded numpy for pyarrow, it broke.

[hmaarrfk]

please do not use sudo and pip at the same time. You very likely broke your system python installation :( That said, I don't think many system utilities use numpy, so you are likely safe.
https://dev.to/elabftw/stop-using-sudo-pip-install-52mn

The installation instructions should be updated with a caution about this
https://scikit-image.org/docs/stable/install.html

The contribution guide goes over how to create virtual environments
https://scikit-image.org/docs/stable/contribute.html#build-environment-setup
Instead of the

pip install -e .

command, use

pip install scikit-image
# or
conda install scikit-image

to install scikit-image in the virtual environment.

[hmaarrfk]

I would try to see if you can install things in a virtual environment, if not, please open a new issue.

I would strongly suggest to use conda for scientific stuff.

[ranka47]

Thanks for the advice.
By removing the dask dependency, it is working.
numpy (1.16.4)
scikit-image (0.14.3)