How to troubleshoot Python installation problems
Run which python3 to see if you are really in the environment you expect to be
Some examples:
$ module purge
$ which python3
/usr/bin/python3 # ← system Python
$ module load python-data
$ which python3
/appl/soft/manual/aida/x86_64/python-data/python-data-2026-03-2/bin/python3
$ module load python-geo
$ which python3
/appl/soft/manual/geo/x86_64/python-geo/3.14.5/bin/python3
$ module load python-pytorch
$ which python3
/appl/soft/manual/aida/x86_64/pytorch/pytorch-2.13-update1/bin/python3
$ module load python-pytorch
$ source venv-pytorch2.13/bin/activate # ← activating virtual environment
$ which python3
/scratch/project_2001234/cscuser/venv-pytorch2.13/bin/python3
You can also check the Python version:
Use pip3 list -v to find out where your packages are coming from
When troubleshooting Python installation problems, we often see that users have installed their own versions of packages that can conflict with those coming from CSC's pre-installed modules. It can be very easy to forget something you installed a long time ago, and it is also easy to miss that pip is installing some additional packages that you are not aware of.
As a reminder see our Python usage guide on how to install your own packages on top of CSC's Python installations.
The command pip3 list -v is an easy way to find out where your
packages are coming from. The list might be very long, so you might
want to use less to get a view where you can scroll up and down, or
grep to extract the lines you are interested in.
In less you can use the arrow keys or Page Up and Page Down keys to
move around. Press 'q' to quit.
Below is an example (with many lines removed) using the python-geo module combined with a venv and some packages installed into the user's home directory.
Package Version Location Installer
------------- ------- ---------------------------------------------------------------------- ---------
geopandas 1.1.3 /ROIHU_TYKKY_euxliVc/miniforge/envs/env1/lib/python3.14/site-packages conda # ← tykky
geoparquet-io 1.2.0 /ROIHU_TYKKY_euxliVc/miniforge/envs/env1/lib/python3.14/site-packages pip # ← tykky
..
biopython 1.88 /users/cscuser/.local/x86_64/lib/python3.14/site-packages pip # ← user's home
..
cowsay 6.1 /projappl/project_2001234/cscuser/my-venv/lib/python3.14/site-packages pip # ← project venv
In the pip3 list -v output, check the Location column. In the example
above the packages with locations starting with /ROIHU_TYKKY_ are
those coming from the python-geo module, which has been installed
with Tykky. Note that these are paths inside the container
created with Tykky. The package biopython seems to be installed into
the user's home directory, while cowsay is in the user's venv
virtual environment found in a project folder.
To see only packages not coming from the CSC-installed python-geo module:
Note that some modules are not using Tykky and Conda, for
example python-pytorch has packages installed into /usr/local/lib or
/usr/local/lib64. Note that these are paths inside the python-pytorch
container.