Closed ahafri closed 6 years ago
Hi @ahafri, this looks like a memory error (potentially duplicate of #1162, #1209, etc.).
First, we just have released 1.1.3 with a bugfix for this. Please check out that new version.
If the error is still happening, please set the flag --mem-mb
to a low value (lower than your actual memory ceiling). For instance, if your machine has 32GB RAM, set that to some 20GB - --mem-mb 20000
.
If this is still happening, then try the flat --low-mem
.
Please let us know if this worked for you.
Hey, I tried 1.1.3 but was having Freesurfer license issues (wasn't detecting my license file) so I stuck with 1.0.14. However, the --mem-mb 20000
worked: no aCompCor
or tCompCor
errors!
Thanks for the update.
I strongly recommend you to update to the latest release (1.1.4), so please open a new issue describing what you tried, and how it doesn't work for you.
Also, please check on neurostars.org for users having similar issues: https://neurostars.org/search?q=license%20tags%3Afmriprep
I have run fmriprep on 7 subjects, each with 3-4 sessions, and 8 or so runs per session.
Out of the 7 subjects, two subjects have had one run each (different runs) in which there are errors for aCompCor and tCompCor, and as a result no confounds.tsv file is generated.
I had gotten this error for a couple more runs than this, but I think it was a memory issue then; when I allocated more RAM for the jobs (24GB), those runs no longer had errors. However, for two runs the aCompCor and tCompCor errors still exist (even without an "Out of Memory" error).
I am using fmriprep version 1.0.14 (May 15, 2018), via a singularity container. The singularity command:
FMRIPrep command (within singularity):
Specific errors are identical and are listed below for one example run: