Excercise 4 code compiling issue; paraview module not found

[ChristianZaeh]

Iḿ using VS code with the jupyter notebook extension. When compiling the code copied from the excercise 4 folder:

from glob import glob
from os.path import join
import torch as pt
import matplotlib.pyplot as plt
from flowtorch.data import FOAMDataloader

#
# adjust the path if necessary
#
cases = glob("./boundary_layer_1D_variation/Ub_*")

cases = sorted(cases, key=lambda case: float(case.split("_")[-1]))
loader = FOAMDataloader(cases[0])
y = loader.vertices[:, 1]
u_x = pt.zeros((y.shape[0], len(cases)))
for i, case in enumerate(cases):
    loader = FOAMDataloader(case)
    u_x[:, i] = loader.load_snapshot("U", loader.write_times[-1])[:, 0]

Ubar = pt.tensor([float(case.split("_")[-1]) for case in cases])
print("Shape of data matrix: ", u_x.shape)

# creating a plot
delta, nu = 0.5, 1.0e-5
Re = pt.tensor([Ub.item()*2*delta/nu for Ub in Ubar])
for i, Ub in enumerate(Ubar):
    plt.plot(u_x[:, i], y, label=r"$Re={:1.0f}$".format(round(Re[i].item(), 0)))
plt.xlabel(r"$u_x$")
plt.ylabel(r"$y$")
plt.xlim(0.0, 1.1)
plt.ylim(-0.01, 0.5)
plt.legend(loc="upper center", ncol=4, bbox_to_anchor=[0.5, 1.3])
plt.show()

I get the following error message:

ModuleNotFoundError                       Traceback (most recent call last)
Cell In [1], line 5
      3 import torch as pt
      4 import matplotlib.pyplot as plt
----> 5 from flowtorch.data import FOAMDataloader
      7 #
      8 # adjust the path if necessary
      9 #
     10 cases = glob("./boundary_layer_1D_variation/Ub_*")

File ~/ml-cfd/lib/python3.10/site-packages/flowtorch/data/__init__.py:7
      5 from .psp_dataloader import PSPDataloader
      6 from .tau_dataloader import TAUDataloader, TAUConfig
----> 7 from .tecplot_dataloader import TecplotDataloader
      8 from .selection_tools import mask_box, mask_sphere
      9 from .outlier_tools import iqr_outlier_replacement

File ~/ml-cfd/lib/python3.10/site-packages/flowtorch/data/tecplot_dataloader.py:10
      8 # third party packages
      9 import torch as pt
---> 10 from paraview import servermanager as sm
     11 from paraview.vtk.numpy_interface import dataset_adapter as dsa
     12 from paraview.simple import VisItTecplotBinaryReader

ModuleNotFoundError: No module named 'paraview'

I already reinstalled the python environment and also paraview and still get the error. I assume my virtual environment doesnt regognize the libraries but I couldnt fix it. Do I have to manually add the paraview libraries to the pythonpath?
Thank you for the help!

[Janis Geise]

Hi @ChristianZaeh,

did you re-installed the flowtorch library as described at the top of lecture 4 + 5 ? You can also check out issue #20 and #25 for further information.

If you already installed the correct version of flowtorch and this didn't solve the issue please let me know.

Regards
Janis

[ChristianZaeh]

Hello @JanisGeise,
no I missed that, but I still get the error.
Regards
Christian

[Janis Geise]

Hi @ChristianZaeh,

in VS code, can you check if you selected the correct interpreter / virtual environment for executing this script? From the error message ~/ml-cfd/lib/python3.10/site-packages/flowtorch/data/tecplot_dataloader.py:10 it looks like you are currently using python3.10 in your IDE. The virtual environment should be python3.8.10.

Another option would be to try executing the script from a terminal using the virtual environment to make sure all dependencies are installed correctly. If the execution from a terminal works, then this should be an issue with the settings in VS code.

Regards,
Janis

[ChristianZaeh]

Hi @JanisGeise,

I installed the virtual environment with sudo apt install python3-venv and python3 -m venv ml-cfd. In this case the 3.10 version gets installed. I thought, according to the tutorial tutorial in excercise 1, that this is not a problem.
When I try to use sudo apt install python3.8-venv I get the errors Unable to locate package python3.8-venv and Couldn't find any package by glob 'python3.8-venv'

Regards,
Christian

[Janis Geise]

ok, that's weird, the script in general should work with python3.10 as well (but there may occur conflicts in future exercises when using python3.10). The only suggestion I have at the moment is to remove and then re-create the ml-cfd environment from scratch. Otherwise, it's probably a good idea to ask @AndreWeiner at tomorrow's exercise.

Regards,
Janis

[Andre Weiner]

Hi @ChristianZaeh,
first, please follow again the instructions provided in the notebook step by step:

# starting from the top-level lecture folder
# enter the Python environment
source ml-cfd/bin/activate
# now you should be in the environment
# remove the old version of flowTorch
pip uninstall flowtorch
# install the latest version
pip install git+https://github.com/FlowModelingControl/flowtorch.git@aweiner

Now, navigate into the notebook folder, open jupyter lab, and open the ml_intro.ipynb notebook:

(ml-cfd) your_name@your_host:~/path/to/ml-cfd-lecture$ cd notebooks
(ml-cfd) your_name@your_host:~/path/to/ml-cfd-lecture$ jupyter-lab

Now the execution should work.

When working in vscode, make sure you are using the correct environment (upper right corner in the picture below). If a different environment is displayed, click on the displayed name and choose ml-cfd (Python...)

Best, Andre