Co-occurrence gives 'nan' if n_splits>1 in large dataset

Open Mena-SA-Kamel opened this issue 2 years ago • 1 comments

I am re-opening this issue because I am still having problems running squidpy.gr.co_occurrence() with n_splits>1. I noticed that this has been previously discussed here and was closed without resolution. When I run on a dataset with 101510 observations while setting n_splits=1, I get an OOM error. If I don't specify an n_splits value, it sets it to 50 and I end up with all NaNs in my occurrence array. Is there a work around that you may suggest?

May 10 '23 17:05 Mena-SA-Kamel

hi @Mena-SA-Kamel , since 1.5 it should have been fixed, could you try and install the latest version of squidpy?

Jul 07 '24 13:07 giovp

I am closing this since there is no response. @Mena-SA-Kamel feel free to re-open if the issue persists

Dec 02 '25 10:12 selmanozleyen