Feature #1336
closed
prepare 3D refinement without cluster specifics and then prepare for parallel processing
Added by Anchi Cheng over 13 years ago.
Updated over 13 years ago.
Description
As part of recon refactoring, all 3D refinement should go through preparation where
- The particle stack may be binned and low-pass filtered without reboxing or being considered as a new stack
- particle stack may need to change format without being considered as a new stack
- initial model rescaled
- core-part of the job script written without cluster specifics
- a list of files to be copied to the cluster should be compiled with subdirectories tared into a file.
This will then allow cluster selection to be done later with the cluster class Christopher is writing.
Once files are copied, in the jobfile at the cluster
- untar files
- prepare parallel refinement script according to number of processor (an variable in the job file).
- run the refinement script
r15706 is the first attempt for Frealign. Still need to figure out database recording part of the refinestack and how to make new non-ctf-corrected stack for this without too much delay of time and still related to the original stack.
Here is an example of what the job script on the cluster would look like.
## frealign_recon6
#PBS -l nodes=1:ppn=8
#PBS -l walltime=24:00:00
#PBS -l cput=240:00:00
#PBS -l mem=2gb
#PBS -m e
#PBS -r n
#PBS -j oe
# refineStackId: 26
# reconStackId:
# modelId: 1
/bin/rm -rf /ami/data00/appion/zz07jul25b/recon/frealign_recon6/recon
mkdir -p /ami/data00/appion/zz07jul25b/recon/frealign_recon6/recon
cd /ami/data00/appion/zz07jul25b/recon/frealign_recon6/recon
#copy job file here for update purpose later
/bin/cp /ami/data00/appion/zz07jul25b/recon/$PBS_JOBNAME .
#download files from main filesystem
ln -s /ami/data00/appion/zz07jul25b/stacks/stack24-meanfit-no-ctf/start.hed start.hed
ln -s /ami/data00/appion/zz07jul25b/stacks/stack24-meanfit-no-ctf/start.img start.img
ln -s /ami/data00/appion/zz07jul25b/recon/frealign_recon6/frealign_recon6.tar frealign_recon6.tar
setenv RUNPAR_RSH 'ssh'
tar -xkf frealign_recon6.tar
echo 'set total number of processors'
setenv TOTAL_PROCS '8'
echo 'preparing Frealign parameters for parallel run'
/opt/myamisnap/bin/appion prepParallelFrealign.py --nproc=$TOTAL_PROCS --jobtype=frealign --recondir=/ami/data00/appion/zz07jul25b/recon/frealign_recon6/recon
echo 'starting particle refinement for iter 1' > refine.log
mpirun --hostfile $PBS_NODEFILE -np $TOTAL_PROCS --app iter001/frealign.iter001.run.sh
echo 'particle refinement complete for iter 1' >> refine.log
echo 'starting volume reconstruction for iter 1' >> refine.log
mpirun --hostfile $PBS_NODEFILE -np 1 iter001/frealign.iter001.combine.sh
test -s threed.001a.img || exit
test -s params.iter001.par || exit
echo 'volume reconstruction complete for iter 1' >> refine.log
echo 'iteration 1 is complete' >> refine.log
echo '' >> refine.log
echo 'starting particle refinement for iter 2' >> refine.log
mpirun --hostfile $PBS_NODEFILE -np $TOTAL_PROCS --app iter002/frealign.iter002.run.sh
echo 'particle refinement complete for iter 2' >> refine.log
echo 'starting volume reconstruction for iter 2' >> refine.log
mpirun --hostfile $PBS_NODEFILE -np 1 iter002/frealign.iter002.combine.sh
test -s threed.002a.img || exit
test -s params.iter002.par || exit
echo 'volume reconstruction complete for iter 2' >> refine.log
echo 'iteration 2 is complete' >> refine.log
echo '' >> refine.log
tar -czf results.tgz *iter* *.txt *.log
tar -czf models.tgz threed*
#upload files back to main filesystem
/bin/mv -v results.tgz /ami/data00/appion/zz07jul25b/recon/frealign_recon6/
/bin/mv -v models.tgz /ami/data00/appion/zz07jul25b/recon/frealign_recon6/
/bin/mv -v $PBS_JOBNAME /
cd /ami/data00/appion/zz07jul25b/recon/frealign_recon6
- Status changed from Assigned to Closed
This idea was not working. replaced with #1342
Also available in: Atom
PDF