UNITO_User_Interface/mainGUI.py generates ImportError, missing PyQt5

[MorselHoarder]

Hi,
Thanks for all your hard work on UNITO! 🙂 Looking forward to trying it out.

Upon a fresh install with anaconda I'm getting an error when running the mainGUI.py module.

$ cd UNITO/UNITO_User_Interface/
$ python mainGUI.py
Traceback (most recent call last):
  File "/home/schult/UNITO/UNITO_User_Interface/mainGUI.py", line 4, in <module>
    from matplotlib.backends.backend_qt5agg import FigureCanvasQTAgg as FigureCanvas
  File "/home/schult/anaconda3/envs/UNITO_demo/lib/python3.9/site-packages/matplotlib/backends/backend_qt5agg.py", line 7, in <module>
    from .backend_qtagg import (    # noqa: F401, E402 # pylint: disable=W0611
  File "/home/schult/anaconda3/envs/UNITO_demo/lib/python3.9/site-packages/matplotlib/backends/backend_qtagg.py", line 9, in <module>
    from .qt_compat import QT_API, QtCore, QtGui
  File "/home/schult/anaconda3/envs/UNITO_demo/lib/python3.9/site-packages/matplotlib/backends/qt_compat.py", line 130, in <module>
    raise ImportError(
ImportError: Failed to import any of the following Qt binding modules: PyQt5, PySide2

To reproduce, I followed the anaconda installation instructions in UNITO_User_Interface/README.md exactly.

$ conda create --name UNITO_demo python=3.9
$ conda activate UNITO_demo
$ conda install pytorch::pytorch torchvision torchaudio -c pytorch
$ conda install anaconda::numpy pandas matplotlib seaborn scipy scikit-learn pyqt
$ conda install -c conda-forge opencv albumentations datashader

I noticed mainGUI.py uses PyQt5, however the most recent version of pyqt installed with conda's default channel is version 6.

$ conda list
...
pyqt           6.7.1           py39h8dad735_2
pyqt6-sip      13.9.1          py39h6ce4db3_2
...

I tried solving this by installing some form of PyQt5 using conda, but the default and anaconda channels don't have a PyQt5-specific package available. I also tried uninstalling pyqt and reinstalling with a pinned version conda install 'pyqt<6', but the python version pinned would require making significant version pins in other packages.

I solved this by running pip install pyqt5 in the UNITO_demo environment. The working conda list looks like this:

$ conda list
pyqt                           6.7.1            py39h8dad735_2
pyqt5                          5.15.11          pypi_0                        pypi
pyqt5-qt5                      5.15.17          pypi_0                        pypi
pyqt5-sip                      12.17.0          pypi_0                        pypi
pyqt6-sip                      13.9.1           py39h6ce4db3_2

While this is a simple fix, I would highly recommend generating some sort of requirements.txt or other lock file with version pins to avoid future package problems.

[KyleeCJ]

Hi, thank you so much for your interest in UNITO and your valuable suggestions! I have uploaded a requirements.txt file in the UNITO_User_Interface folder so that users can use it as a reference for potential version conflicts. Thanks!

[MorselHoarder]

Thank you!! This is extremely helpful! 😄

May I make another request that you use conda-project and export an environment.yml for each of the UNITO/UNITO_* packages? I think that would be the most helpful for people trying to reproduce the environment (I don't have a mac). It's pretty easy and docs can be found here: https://www.anaconda.com/docs/getting-started/working-with-conda/environments#locking-an-environment

[KyleeCJ]

Hi,

Thank you so much for your suggestions! I have added both requirements.txt file and environment.yml files in each folder. Thanks!