| Author |
 Topic  |
|
|
Dichter
BIAC Faculty
190 Posts |
 Posted - Feb 26 2013 : 09:13:04 AM
|
I am trying to running the following model on Hugin with FLAME 1:
http://can.unc.edu/images/uploads/design.png
This model has been running for 24 hours and appears to be only 1/3 finished so far. Any recommendations for how to speed this up? Is it possible to take advantage of multiple nodes for a single FEAT job? Thanks, |
Gabriel S. Dichter, PhD
UNC Departments of Psychiatry & Psychology
http://www.can.unc.edu/ |
|
|
petty
BIAC Staff
USA
453 Posts |
Posted - Feb 26 2013 : 09:43:12 AM
|
The cluster is set-up to take advantage of some pre-programmed SGE submissions.
To turn it on, export the environmental variable FSLSUBMIT=1 ( i've defaulted to 0, which is off ).
Some parts of your feat model will be automatically submitted, then put back together at the end. However, this only might cut your time in half, since you really on have 2 copes |
 |
|
|
syam.gadde
BIAC Staff
USA
421 Posts |
Posted - Feb 26 2013 : 09:45:53 AM
|
Although the docs on the website say that FLAME will do slices in parallel, so maybe it will speed things up more.
In case it wasn't clear, you put:
export FSLSUBMIT=1
somewhere in your submission script before the call to feat or if you're running interactively run that at the command line before running FLAME. |
|
|
|
Dichter
BIAC Faculty
190 Posts |
Posted - Feb 26 2013 : 09:48:23 AM
|
Thank you both. Any way to use that environmental variable when running from the GUI?
Incidentally, removing the individual subject mean EVs (EV3-E39) and replacing with a grand mean variable (all 1's), which *should* yield similar result cuts the processing time down considerably. |
Gabriel S. Dichter, PhD
UNC Departments of Psychiatry & Psychology
http://www.can.unc.edu/ |
|
|
|
syam.gadde
BIAC Staff
USA
421 Posts |
Posted - Feb 26 2013 : 09:50:08 AM
|
Same thing, run:
export FSLSUBMIT=1
before running the GUI. |
|
|
|
Dichter
BIAC Faculty
190 Posts |
Posted - Feb 26 2013 : 8:50:02 PM
|
| Job finished in ~6 hours with this option. Thank you! |
Gabriel S. Dichter, PhD
UNC Departments of Psychiatry & Psychology
http://www.can.unc.edu/ |
|
|
| |
Topic |
|
Processing time for FEAT model
