Closed clhinrichs closed 1 year ago
Hi! Thank you for the feedback.
Regarding your issues:
Yes, "environment-cuda.yml" contains the list of packages which are needed to run the code on the GPU nodes
Part of the README was not up to date, in particular regarding the names of the input directories which have been recently changed. The documentation is now updated. However, it is not needed to copy the input netCDF files to run the demo. Indeed, data-root-dir in "demo_args.txt" is already pointing to a directory containing the input file. It is only required to execute crai-evaluate --load-from-file demo_args.txt
in the demo directory.
This message is a warning message which seems to be related to the torchvision version (see discussion). It has no impact on the code itself and especially not on the creation of the png file when running the demo. This file is created in the "outputs" directory and is named "demo_combined.1_0.png".
Please, let us know if the updated README is helping you to generate the png file.
Thank you for the clarification and the updated readme. I could execute the demo case with no further problems and ignored the warning re torchvision. Thanks for your support!
Hi, I cloned and installed crai on Levante and I am trying to run the demo. A few questions / issues came up:
1) On Levante should I use the crai-cuda environment when I planning to use GPUs? 2) In demo, I do not have the inputs/test-large folder. I copied the tas* file from data/test for now 3) Then I get this message after crai-evaluate --load-from-file demo_args.txt /home/a/a270109/.conda/envs/crai-cuda/lib/python3.10/site-packages/torchvision/io/image.py:13: UserWarning: Failed to load image Python extension: libjpeg.so.8: cannot open shared object file: No such file or directory warn(f"Failed to load image Python extension: {e}")
as output I get nc files but not the png's.
Thanks for your help!