CPU for running nilearn SVC on ~67 subjects




I have attempted to run a decoding analysis with a linear SVC in 67 people (475 measurements each). I’m running on a HPC with as much as 128GB memory, but my job keeps hitting the memory cap and failing. This happens specifically at the SVC step.
–> output: https://github.com/jennyrgilb/choc_nilearn/blob/master/imagine_nilearn_half.output

How much memory should this take? I’m worried that something in the script it hanging, causing it to eat up all the memory, and the job is killed. This is half of the total sample I’d like to run.

My script is here: https://github.com/jennyrgilb/choc_nilearn/blob/master/img_nested_app_v_unapp_half.py

Thanks in advance for your help!