Dear all,
Just to pick up on Leonid Sazanov's post on October 27th:
"Also, there seems to be a new bug in the GUI for movie refinement job -
the second part of the job (movie refinement) tries to start straight
away, before the first part (movie particle extraction) is finished and
so the whole job crashes due to lack of movie particles. We work around
that by submitting two separate command line jobs, but would be good to
fix this in GUI."
We too note this issue for Relion 2.1-b1 when trying to run on a GPU
cluster. For whatever reason, much of the GUI input is not translated
to a full command. Here's the command actually issued instead:
mpirun -N 5 -n 5
/usr/local/Cluster-Apps/relion-gpu/2.1.b1-wilkes-2/bin/relion_refine_mpi
--o MovieRefine/job132/run_ravg7_off1 --continue
./Refine3D/job064/run_it014_optimiser.star --process_movies_in_batches
--realign_movie_frames MovieRefine/job132/micrographs_movie_list.star
--movie_frames_running_avg 7 --movie_name movie --sigma_off 1
--skip_rotate --skip_maximize --j 1
This immediately fails as it tries to look for
micrographs_movie_list.star in the new job folder (job132), where it
doesn't reside.
Interestingly, running with the exact same GUI settings and data from
Relion 2.0 generates a much fuller command:
mpirun -ppn 3 -np 3
/usr/local/Cluster-Apps/relion-gpu/2.0/bin/relion_preprocess_mpi --i
./MotionCorr/job004/corrected_micrograph_movies.star
--reextract_data_star Refine3D/job064/run_it014_data.star --part_dir
MovieRefine/job090/ --list_star
MovieRefine/job090/micrographs_movie_list.star --join_nr_mics 50
--extract --extract_movies --extract_size 284 --movie_name movie
--first_movie_frame 2 --last_movie_frame 0 --avg_movie_frames 1 --scale
142 --norm --bg_radius 53 --white_dust -1 --black_dust -1 --invert_contrast
(note, this is necessarily running on a different cluster that uses
different MPI libraries, hence the different MPI flags - however, all
other processes tried so far with both 2.0 and 2.1-b1 work absolutely fine).
The extra challenge I have is that even the 2.0 job does not run to
completion. The GUI-issued command does the initial re-extraction but
then does not launch the second process of refinement. Even trying to
use Continue does not get the second process running - the log file
notes that extraction is skipped (as the program can see it has already
been done), but then nothing else happens. There are no error messages
or unusual exit codes that I can see.
In short I can't get movie refinement to run with either 2.0 or 2.1-b1.
Short of trying command-line job submission, does anyone have any
potential solutions?
Many thanks,
Neil
--
Dr. Neil Rzechorzek
Department of Biochemistry
University of Cambridge
Sanger Building
80 Tennis Court Road
Cambridge, UK
CB2 1GA
www.bioc.cam.ac.uk
Advisory Chairman - Entrepreneurial Postdocs Of Cambridge
www.epoc.group.cam.ac.uk
Email: [log in to unmask]
Tel: +44 (0)1223 333652
|