def run_lsf_blat(subject,queries,blattile,blatargstr='',num_batches=100):
'''submits mcl_id_triples_by_blat.py jobs to LSF
intended as an example of parallelization over a compute grid;
uses a module LSF.py for interaction with scheduler
'''
import LSF
blatargstr += ' -tileSize=%s' % blattile
blatargstr += ' -stepSize=%s' % (int(blattile)/2)
cmds = []
labf = []
for q in queries:
outbase = q.rstrip('.fa').rstrip('_query')+'_blat'+blatargstr.replace('=','').replace(' ','')
labf.append(outbase+'.label')
cmds.append('%smcl_id_triples_by_blat.py %s %s \\"%s\\" %s' % (radtag_denovo,subject,q,blatargstr,outbase))
logfile = os.path.join(os.path.dirname(subject),'blat-log')
try:
os.unlink(logfile)
except:
pass
#print >> sys.stderr, 'LSF %s\nlog: %s' % (cmds,logfile)
import time
while len(cmds) > 0:
jobids,namedict = LSF.lsf_jobs_submit(cmds,logfile,'normal_serial',jobname_base='blat2mat',num_batches=num_batches)
time.sleep(20)
LSF.lsf_wait_for_jobs(jobids,logfile,namedict=namedict)
cmds = LSF.lsf_no_success_from_log(logfile)
return labf
def run_lsf_blat(subjects,queries,blattile,blatargstr='',num_batches=100):
'''submits mcl_id_triples_by_blat.py jobs to LSF
intended as an example of parallelization over a compute grid;
uses a module LSF.py for interaction with scheduler
'''
import LSF
blatargstr += ' -tileSize=%s' % blattile
blatargstr += ' -stepSize=%s' % (int(blattile)/2)
cmds = []
labf = []
for q in queries:
for subject in subjects:
subjname = os.path.basename(subject).rstrip('.fa').rstrip('_subj')
outbase = q.rstrip('.fa').rstrip('_query')+'_blat'+'-subj'+subjname+blatargstr.replace('=','').replace(' ','')
labf.append(outbase+'.label')
cmds.append('%smcl_id_triples_by_blat.py %s %s \\"%s\\" %s' % (radtag_denovo,subject,q,blatargstr,outbase))
logfile = os.path.join(os.path.dirname(subjects[0]),'blat-log')
logfiles = glob(logfile+'*.lsflog')
for lf in logfiles:
try:
os.unlink(lf)
except:
pass
#print >> sys.stderr, 'LSF %s\nlog: %s' % (cmds,logfile)
import time
while len(cmds) > 0:
jobids,namedict = LSF.lsf_jobs_submit(cmds,logfile,'normal_serial',bsub_flags='-R "select[mem > 20000]"',jobname_base='blat2mat',num_batches=num_batches)
time.sleep(20)
LSF.lsf_wait_for_jobs(jobids,logfile,namedict=namedict)
logfiles = glob(logfile+'*.lsflog')
cmds = reduce(lambda x,y:x+y, [LSF.lsf_no_success_from_log(lf) for lf in logfiles])
if not all([os.path.exists(f) for f in labf]):
raise OSError, 'blat failed'
return labf
def run_parallel_blasts(clids,mapname,gr,tab,uniqued,blastdb,grname=None,nbatches = 10):
if grname is None:
grname = os.path.basename(gr)
batchlen = len(clids) / nbatches
clid_batches = [clids[i:i+batchlen] for i in range(0,len(clids),batchlen)]
cmds = []
blastouts = []
for cb in clid_batches:
seqfile = os.path.join(outroot,'%s_%s-%s_%sclust_%s.fa' % (os.path.splitext(mapname)[0],cb[0],cb[-1],len(cb),os.path.splitext(grname)[0]))
blast_out = os.path.join(outroot,'%s-%s.blast' % tuple([os.path.splitext(os.path.basename(f))[0] for f in [seqfile,blastdb]]))
blastouts.append(blast_out)
if os.path.exists(blast_out): continue
clidlist = '\n'.join(cb)
cmds.append('echo -e "%s" | blast_from_rad_clusters.py %s %s %s %s %s %s' % (clidlist,gr,tab,uniqued,seqfile,blastdb,blast_out))
logfile = os.path.join(outroot,'rad-clust-blast-log')
jobids,namedict = LSF.lsf_jobs_submit(cmds,logfile,'normal_serial',bsub_flags='-R "select[mem>30000]"', jobname_base='radclustblast',num_batches=nbatches)
LSF.lsf_wait_for_jobs(jobids,logfile,namedict=namedict,restart_z=24)
return blastouts
cmds = [] rerun = True while rerun: for clab,crops in cropsdict.items(): outbase,outext = os.path.splitext(vid) outvid = '%s_%s_%s-%s%s' % (outbase,clab,offset,dur,outext) if os.path.exists(outvid) and ( vidtools.vid_duration(outvid) == dur ): print >> sys.stderr, '%s present and expected size, skip' % outvid else: cropstr = '-vf crop=in_w-%s:in_h-%s:%s:%s' % (crops[0]+crops[2],crops[1]+crops[3],crops[0],crops[1]) cmd = 'ffmpeg -ss %s -t %s -i %s -y %s -b 20000k %s' % (offset,dur,vid,cropstr,outvid) cmds.append(cmd) logfile = os.path.join(os.path.dirname(vid),'crop-log') jobids,namedict = LSF.lsf_jobs_submit(cmds,logfile,'normal_serial',jobname_base='vid2crop') LSF.lsf_wait_for_jobs(jobids,logfile,'normal_serial',namedict=namedict) cmds = LSF.lsf_no_success_from_log(logfile) if len(cmds) == 0: rerun = False sys.exit() rerun = True while rerun: outstr = os.path.join(outroot,'%07d.png') execstr = 'ffmpeg -ss %s -t %s -i %s -r %s -y %s %s 2> /dev/null' % (offset,dur,vid,fps,cropstr,outstr) print >> sys.stderr, 'execute %s\nrunning %s' % (nrun,execstr) os.system(execstr) nrun += 1
new_tdirs = vidtools.parallel_v2p(vid, fps,tdir=source_root,queue='short_serial',num_jobs=20,crops=crops,extract_windows=nonex_wins)
tdirs = exist_paths + new_tdirs
cmds = ['reanalyze_activity_segments.py %s %s %s' % (source_adir, d, new_seglen) for d in tdirs]
logfile = source_root+'/reanalyze-log'
final_summary = source_root+'/%0.1fsec_summary.pdf' % new_seglen
if os.path.exists(final_summary):
print >>sys.stderr, 'merged summary %s exists; will not reanalyze' % final_summary
do = False
else:
do = True
passes = 0
while do and passes < 3:
jids,ndict = LSF.lsf_jobs_submit(cmds,logfile,'normal_serial',jobname_base='reanalyze')
LSF.lsf_wait_for_jobs(jids,logfile,ndict)
time.sleep(10)
unfinished = LSF.lsf_no_success_from_log(logfile)
if unfinished:
print >> sys.stderr, 'not finished: %s' % unfinished
cmds = unfinished
else:
do = False
passes += 1
if not os.path.exists(final_summary):
print >> sys.stderr, 'write summary to '+final_summary
os.system('gs -dBATCH -dNOPAUSE -q -sDEVICE=pdfwrite -sOutputFile=%s %s/*-*/30fps/analysis/1.0sec*/summary.pdf' % (final_summary,source_root))
#!/usr/bin/env python
import os,sys,LSF
from PIL import Image
from glob import glob
if len(sys.argv[1:]) > 1 or sys.argv[1].endswith('*.png'):
for f in sys.argv[1:]:
im = Image.open(f)
try:
h=im.histogram()
except:
print 'invalid image: ' + f
os.unlink(f)
else:
outfile = sys.argv[1]+'/../drop-broken-pngs-log'
print >> sys.stderr, 'master running, target dir %s output in %s' % (sys.argv[1],outfile)
cmds = []
images = glob(sys.argv[1]+'/*.png')
for i in range(0,len(images),1000):
cmds.append(sys.argv[0]+' '+(' '.join(images[i:i+1000])))
jids,ndict = LSF.lsf_jobs_submit(cmds,outfile,jobname_base='pngdrop',num_batches=400)
LSF.lsf_wait_for_jobs(jids,restart_outfile=outfile,namedict=ndict,restart_z=12)
restart_z = 12
while rerun:
cmds = []
for i in range(0,len(images),int(segment_step)):
this_out = os.path.join( outroot,'%07d-%07d.mice' % (i,i+segment_step))
if not os.path.exists(this_out):
cmd = 'summarize_segment.py -m %s -s %d -e %d -p %d -t %d -r %s -o %s -x \\"%s\\" -b \\"%s\\" -g \\"%s\\" %s' \
% (mask,i,i+segment_step,pixav,timeav,outroot,mousez,opts['xybounds'],opts['burrow_entrance_xy'],opts['ground_anchors'],imagedir)
cmds.append(cmd)
# drop last command (incomplete segment)
dropcmd = cmds.pop()
logfile = os.path.join(outroot,'summarize-segment-log')
print >> sys.stderr,'running summary of %s segments, log written to %s\n' % (len(cmds),logfile)
print >> sys.stderr,'bundle into %s batches' % num_batches
jids,ndict = LSF.lsf_jobs_submit(cmds,logfile,opts['summarize_queue'],jobname_base='summarize',num_batches=num_batches,bsub_flags='-R "select[mem > 30000]"')
jobids.update(jids)
namedict.update(ndict)
if glob(outroot+'/*.mice'):
restart_z = None
else:
restart_z = 12
LSF.lsf_wait_for_jobs(jobids,os.path.join(outroot,'restarts'),namedict=namedict,restart_z=restart_z) #restart_z=None)
jobids = {}
namedict = {}
#remove .mice corresponding to mis-sized frames
for f in glob(outroot+'/*.frame'):
fgrp = os.path.splitext(f)[0]
if os.path.getsize(fgrp+'.frame') != 8 * SHAPE[0] * SHAPE[1]:
print >> sys.stderr, 'remove missized file %s (obs: %s exp %s)' % (f, os.path.getsize(fgrp+'.frame'), 8 * SHAPE[0] * SHAPE[1])
'reanalyze_activity_segments.py %s %s %s' % (source_adir, d, new_seglen)
for d in tdirs
]
logfile = source_root + '/reanalyze-log'
final_summary = source_root + '/%0.1fsec_summary.pdf' % new_seglen
if os.path.exists(final_summary):
print >> sys.stderr, 'merged summary %s exists; will not reanalyze' % final_summary
do = False
else:
do = True
passes = 0
while do and passes < 3:
jids, ndict = LSF.lsf_jobs_submit(cmds,
logfile,
'normal_serial',
jobname_base='reanalyze')
LSF.lsf_wait_for_jobs(jids, logfile, ndict)
time.sleep(10)
unfinished = LSF.lsf_no_success_from_log(logfile)
if unfinished:
print >> sys.stderr, 'not finished: %s' % unfinished
cmds = unfinished
else:
do = False
passes += 1
if not os.path.exists(final_summary):
print >> sys.stderr, 'write summary to ' + final_summary
os.system(
'gs -dBATCH -dNOPAUSE -q -sDEVICE=pdfwrite -sOutputFile=%s %s/*-*/30fps/analysis/1.0sec*/summary.pdf'
opts["burrow_entrance_xy"],
opts["ground_anchors"],
imagedir,
)
)
cmds.append(cmd)
# drop last command (incomplete segment)
dropcmd = cmds.pop()
logfile = os.path.join(outroot, "summarize-segment-log")
print >>sys.stderr, "running summary of %s segments, log written to %s\n" % (len(cmds), logfile)
print >>sys.stderr, "bundle into %s batches" % num_batches
jids, ndict = LSF.lsf_jobs_submit(
cmds,
logfile,
opts["summarize_queue"],
jobname_base="summarize",
num_batches=num_batches,
bsub_flags='-R "select[mem > 30000]"',
)
jobids.update(jids)
namedict.update(ndict)
if glob(outroot + "/*.mice"):
restart_z = None
else:
restart_z = 12
LSF.lsf_wait_for_jobs(
jobids, os.path.join(outroot, "restarts"), namedict=namedict, restart_z=restart_z
) # restart_z=None)
jobids = {}
namedict = {}
while rerun:
for clab, crops in cropsdict.items():
outbase, outext = os.path.splitext(vid)
outvid = '%s_%s_%s-%s%s' % (outbase, clab, offset, dur, outext)
if os.path.exists(outvid) and (vidtools.vid_duration(outvid) == dur):
print >> sys.stderr, '%s present and expected size, skip' % outvid
else:
cropstr = '-vf crop=in_w-%s:in_h-%s:%s:%s' % (
crops[0] + crops[2], crops[1] + crops[3], crops[0], crops[1])
cmd = 'ffmpeg -ss %s -t %s -i %s -y %s -b 20000k %s' % (
offset, dur, vid, cropstr, outvid)
cmds.append(cmd)
logfile = os.path.join(os.path.dirname(vid), 'crop-log')
jobids, namedict = LSF.lsf_jobs_submit(cmds,
logfile,
'normal_serial',
jobname_base='vid2crop')
LSF.lsf_wait_for_jobs(jobids, logfile, 'normal_serial', namedict=namedict)
cmds = LSF.lsf_no_success_from_log(logfile)
if len(cmds) == 0:
rerun = False
sys.exit()
rerun = True
while rerun:
outstr = os.path.join(outroot, '%07d.png')
execstr = 'ffmpeg -ss %s -t %s -i %s -r %s -y %s %s 2> /dev/null' % (
offset, dur, vid, fps, cropstr, outstr)
print >> sys.stderr, 'execute %s\nrunning %s' % (nrun, execstr)