AndreWeiner / ml-cfd-lecture

Lecture material for machine learning applied to computational fluid mechanics

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Current ParaView version not compatible with FOAMDataloader

JanisGeise opened this issue · comments

Hi @AndreWeiner,

just a heads up for the upcoming exercises:

The version of ParaView (v5.7) installed in exercise 1 uses python 3.7, but the virtual environment is for python3.8. When importing the flowtorch FOAMDataloader, this will give the following error:

Traceback (most recent call last):
File "/home/.../dataloader.py", line 7, in <module>
from flowtorch.data import FOAMDataloader
File "/home/janis/.local/lib/python3.8/site-packages/flowtorch/data/__init__.py"
from .tecplot_dataloader import TecplotDataloader
File "/home/janis/.local/lib/python3.8/site-packages/flowtorch/data/tecplot_dataloader.py", line 10, in <module>
from paraview import servermanager as sm
ModuleNotFoundError: No module named 'paraview'

The issue can be resolved by installing ParaView version v.5.11 as described in the flowtorch repo. When executing python scripts inside an IDE, the paths added to the ~\.bashrc need to be set in the script before importing flowtorch, e.g. as:

import sys
sys.path.insert(0, "/opt/ParaView-5.9.1-MPI-Linux-Python3.8-64bit/bin")
sys.path.insert(0, "/opt/ParaView-5.9.1-MPI-Linux-Python3.8-64bit/lib/python3.8/site-packages")
from flowtorch.data import FOAMDataloader

Regards, Janis

Hi Janis,
thanks for the heads-up. I already fixed this issue on the flowTorch side by making the import conditional (commit). However, in the requirements.txt, the last flowTorch version from the main branch is installed, where the fix is not yet included. My preferred solution would be to re-install flowTorch in the virtual environment rather than updating ParaView. I'll add a note to both the next exercise and lecture notebooks.
Thanks again!
Andre