Fmriprep memory issues

1

Hi!

I’m trying to run fmriprep 20.2.1 on MAC Mojave (processor 3,2 GHz INtel Xeon W, Memory 32 GB 2666 Mhx DDR) for a dataset with one session per participants and 3 runs per session.
For the docker resources I specified: CPUs 12, Memory 31 GB, Swap 2 GB, Disk image size 152 GB
I use the following flags for fmriprep --low-mem --nprocs 12 --omp-nthreads 8 --mem 30000

And I still get a warning: Some nodes exceed the total amount of memory available (30.00 GB).
And then I quickly get an error: concurrent.futures.process.BrokenProcessPool: A process in the process pool was terminated abruptly while the future was running.

I’ve preprocessed 2 datasets on this machine and everything worked fine.

I tried suggestions from here: Broken process pool · Issue #1260 · nipreps/fmriprep · GitHub
But it doesn’t seem to work.

Do you have any other suggestions or should I find a machine with more RAM?

2

How are you running fMRIprep (conda, Docker, singularity)? Are running on individual subjects or all of them in one command? Can you provide your full command that calls fmriprep?

3

With docker via fmriprep-docker wrapper.
For the docker resources I specified: CPUs 12, Memory 31 GB, Swap 2 GB, Disk image size 152 GB

4

Can you post your full fmriprep command?

5

fmriprep-docker /Desktop/nifti /Desktop/nifti/derivatives participant --participant-label 1 --fs-license /Desktop/license.txt --use-aroma --low-mem --nprocs 12 --omp-nthreads 8 --mem 30000