Skip to content

Massive RAM spike with pyAFQ 3 #218

Description

@nwellman13

Hi there,

This was a something I asked about at Neurohack this year but the proposed solutions didn't seem to help. With version 3 there is an accumulation of RAM after each subject that causes my run to be killed after about 10 subjects with 128 GB RAM. For example, after the first subject completes it uses ~10 GB RAM, then when the next subject completes it's at ~20 GB, etc until it maxes out. This doesn't make sense to me since the previous subjects have completed and should no longer be using memory. It does the same thing whether I provide the segmentations and brain mask or let it create them from the T1. This was not an issue in previous version.

The recommendations I was given was to use either use separate participant labels for each subject in myAFQ object (which I've tried) or to create separate job scripts for each subject and submit to an HPC (which I'm unable to do since I don't currently have access to an HPC). Is this expected behavior of version 3 and if so why? Or could this be a bug? At this point I'm limited to running ~10 subjects at a time with the default bundles until it crashes.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions