Open tkmamidi opened 1 year ago
Something to consider while getting the number of CPUs available for n_workers
- https://stackoverflow.com/a/55423170
Probably a good idea to use os.sched_getaffinity
on linux systems. We'll look into adding that for the next release.
I'm working with slurm on a HPC and couldn't get the
--mp
parameter working. Here's the command I used and the error:oc run testing_variants_hg38.vcf.gz -l hg38 --mp 20 -t csv --package mypackage
Also, it is picking the available CPUs on the node and not from the slurm job resources. maybe the
num_workers
should be modified to reflect situations with HPC usage?Any help is appreciated!
Thanks!