Closed alexlenail closed 12 months ago
I've ran into a similar issue on pySCENIC. I believe it has to do with how each version (pySCENIC & SCENIC+) parallelize their respective binarization functions.
In my case, I was binarizing an auc matrix downstream in pySCENIC. With that versions' binarize function, you can declare number of workers with the pool function from python's multiprocessing package.
To troubleshoot, if you have the option to set this, try not declaring a number of workers. This could force a default based on your environment. If it still doesn't work after that you could check your environments' versions of the packages. (similar to #305)
In my case running things via jupyter notebook, it seemed to just be a kernel issue, and restarting the function after interrupting the run giving the Numba: forking error allowed it to run to completion.
All this with a grain of salt though, since it was with pySCENIC. However I hope it helps you out somewhat!
Thanks Davin! I managed to get it to work in an older conda env. The suggestion to try without multiprocessing workers seems like a good one for future visitors of this thread.
I believe I'm in the step "Binarizing eGRNs AUC" but getting these warnings and errors. The job is still running, but looking at
top
there doesn't seem to be any computing happening.(full log in the details)
I noticed a similar issue was reported on pySCENIC (https://github.com/aertslab/pySCENIC/issues/305), with suggestion this has to do with package version issues.
What do you recommend?