def run(config_file, subject_list_file, p_name = None):
try:
if not os.path.exists(config_file):
raise IOError
else:
c = Configuration(yaml.load(open(os.path.realpath(config_file), 'r')))
except IOError:
print "config file %s doesn't exist" %config_file
raise
except Exception:
raise Exception("Error reading config file - %s"%config_file)
#do some validation
validate(c)
try:
sublist = yaml.load(open(os.path.realpath(subject_list_file), 'r'))
except:
raise Exception ("Subject list is not in proper YAML format. Please check your file")
strategies = sorted(build_strategies(c))
print "strategies ---> ", strategies
sub_ids =[]
for sub in sublist:
if sub['unique_id']:
sub_ids.append(sub['subject_id']+"_" + sub["unique_id"])
else:
sub_ids.append(sub['subject_id'])
create_group_log_template(sub_ids, os.path.join(c.outputDirectory, 'logs'))
seeds_created = []
if not (c.seedSpecificationFile is None):
try:
if os.path.exists(c.seedSpecificationFile):
seeds_created = create_seeds_(c.seedOutputLocation, c.seedSpecificationFile, c.FSLDIR)
print 'seeds created %s -> ' % seeds_created
except:
raise IOError('Problem in seedSpecificationFile')
if 1 in c.runVoxelTimeseries:
if 2 in c.useSeedInAnalysis:
c.maskSpecificationFile = append_seeds_to_file(c.workingDirectory, seeds_created, c.maskSpecificationFile)
if 1 in c.runROITimeseries:
if 1 in c.useSeedInAnalysis:
c.roiSpecificationFile = append_seeds_to_file(c.workingDirectory, seeds_created, c.roiSpecificationFile)
if 1 in c.runNetworkCentrality:
if 3 in c.useSeedInAnalysis:
c.templateSpecificationFile = append_seeds_to_file(c.workingDirectory, seeds_created, c.templateSpecificationFile)
if not c.runOnGrid:
from CPAC.pipeline.cpac_pipeline import prep_workflow
procss = [Process(target=prep_workflow, args=(sub, c, strategies, p_name)) for sub in sublist]
pid = open(os.path.join(c.outputDirectory, 'pid.txt'), 'w')
import subprocess
jobQueue = []
if len(sublist) <= c.numSubjectsAtOnce:
"""
Stream all the subjects as sublist is
less than or equal to the number of
subjects that need to run
"""
for p in procss:
p.start()
print >>pid,p.pid
else:
"""
Stream the subject worlflows for preprocessing.
At Any time in the pipeline c.numSubjectsAtOnce
will run, unless the number remaining is less than
the value of the parameter stated above
"""
idx = 0
while(idx < len(sublist)):
if len(jobQueue) == 0 and idx == 0:
idc = idx
for p in procss[idc: idc + c.numSubjectsAtOnce]:
p.start()
print >>pid,p.pid
jobQueue.append(p)
idx += 1
else:
for job in jobQueue:
if not job.is_alive():
print 'found dead job ', job
loc = jobQueue.index(job)
del jobQueue[loc]
procss[idx].start()
jobQueue.append(procss[idx])
idx += 1
pid.close()
else:
import commands
import pickle
from time import strftime
temp_files_dir = os.path.join(os.getcwd(), 'cluster_temp_files')
print commands.getoutput("mkdir -p %s" % temp_files_dir)
strategies_file = os.path.join(temp_files_dir, 'strategies.obj')
f = open(strategies_file, 'w')
pickle.dump(strategies, f)
f.close()
if 'sge' in c.resourceManager.lower():
run_sge_jobs(c, config_file, strategies_file, subject_list_file, p_name)
elif 'pbs' in c.resourceManager.lower():
run_pbs_jobs(c, config_file, strategies_file, subject_list_file, p_name)
elif 'condor' in c.resourceManager.lower():
run_condor_jobs(c, config_file, strategies_file, subject_list_file, p_name)
return 1
def run(config_file, subject_list_file, p_name=None, plugin=None,
plugin_args=None):
'''
'''
# Import packages
import commands
import os
import pickle
import time
from CPAC.pipeline.cpac_pipeline import prep_workflow
# Init variables
config_file = os.path.realpath(config_file)
subject_list_file = os.path.realpath(subject_list_file)
# take date+time stamp for run identification purposes
unique_pipeline_id = strftime("%Y%m%d%H%M%S")
pipeline_start_stamp = strftime("%Y-%m-%d_%H:%M:%S")
# Load in pipeline config file
try:
if not os.path.exists(config_file):
raise IOError
else:
c = Configuration(yaml.load(open(config_file, 'r')))
except IOError:
print "config file %s doesn't exist" % config_file
raise
except Exception:
print "Error reading config file - %s" % config_file
raise Exception
# Do some validation
validate(c)
# Get the pipeline name
p_name = p_name or c.pipelineName
# Load in subject list
try:
sublist = yaml.load(open(subject_list_file, 'r'))
except:
print "Subject list is not in proper YAML format. Please check " \
"your file"
raise Exception
# NOTE: strategies list is only needed in cpac_pipeline prep_workflow for
# creating symlinks
strategies = sorted(build_strategies(c))
# Populate subject scan map
sub_scan_map ={}
try:
for sub in sublist:
if sub['unique_id']:
s = sub['subject_id'] + "_" + sub["unique_id"]
else:
s = sub['subject_id']
scan_ids = ['scan_anat']
try:
for id in sub['func']:
scan_ids.append('scan_'+ str(id))
except KeyError:
for id in sub['rest']:
scan_ids.append('scan_'+ str(id))
sub_scan_map[s] = scan_ids
except:
print "\n\n" + "ERROR: Subject list file not in proper format - " \
"check if you loaded the correct file?" + "\n" + \
"Error name: cpac_runner_0001" + "\n\n"
raise Exception
create_group_log_template(sub_scan_map, c.logDirectory)
pipeline_timing_info = []
pipeline_timing_info.append(unique_pipeline_id)
pipeline_timing_info.append(pipeline_start_stamp)
pipeline_timing_info.append(len(sublist))
# If we're running on cluster, execute job scheduler
if c.runOnGrid:
# Create cluster log dir
cluster_files_dir = os.path.join(c.logDirectory, 'cluster_files')
if not os.path.exists(cluster_files_dir):
os.makedirs(cluster_files_dir)
# Create strategies file
strategies_file = os.path.join(cluster_files_dir, 'strategies.obj')
with open(strategies_file, 'w') as f:
pickle.dump(strategies, f)
# Check if its a condor job, and run that
if 'condor' in c.resourceManager.lower():
run_condor_jobs(c, config_file, strategies_file,
subject_list_file, p_name)
# All other schedulers are supported
else:
run_cpac_on_cluster(config_file, subject_list_file,
strategies_file, cluster_files_dir)
# Run on one computer
else:
# Init variables
procss = [Process(target=prep_workflow,
args=(sub, c, strategies, 1, pipeline_timing_info,
p_name, plugin, plugin_args))
for sub in sublist]
if not os.path.exists(c.workingDirectory):
try:
os.makedirs(c.workingDirectory)
except:
err = "\n\n[!] CPAC says: Could not create the working " \
"directory: %s\n\nMake sure you have permissions " \
"to write to this directory.\n\n" % c.workingDirectory
raise Exception(err)
pid = open(os.path.join(c.workingDirectory, 'pid.txt'), 'w')
# Init job queue
jobQueue = []
# If we're allocating more processes than are subjects, run them all
if len(sublist) <= c.numParticipantsAtOnce:
for p in procss:
p.start()
print >>pid, p.pid
# Otherwise manage resources to run processes incrementally
else:
idx = 0
while idx < len(sublist):
# If the job queue is empty and we haven't started indexing
if len(jobQueue) == 0 and idx == 0:
# Init subject process index
idc = idx
# Launch processes (one for each subject)
for p in procss[idc: idc+c.numParticipantsAtOnce]:
p.start()
print >>pid, p.pid
jobQueue.append(p)
idx += 1
# Otherwise, jobs are running - check them
else:
# Check every job in the queue's status
for job in jobQueue:
# If the job is not alive
if not job.is_alive():
# Find job and delete it from queue
print 'found dead job ', job
loc = jobQueue.index(job)
del jobQueue[loc]
# ...and start the next available process
# (subject)
procss[idx].start()
# Append this to job queue and increment index
jobQueue.append(procss[idx])
idx += 1
# Add sleep so while loop isn't consuming 100% of CPU
time.sleep(2)
# Close PID txt file to indicate finish
pid.close()
def run(config_file, subject_list_file, p_name = None):
# Import packages
import time
# take date+time stamp for run identification purposes
unique_pipeline_id = strftime("%Y%m%d%H%M%S")
pipeline_start_stamp = strftime("%Y-%m-%d_%H:%M:%S")
try:
if not os.path.exists(config_file):
raise IOError
else:
c = Configuration(yaml.load(open(os.path.realpath(config_file), 'r')))
except IOError:
print("config file %s doesn't exist" % config_file)
raise
except Exception:
print("Error reading config file - %s" % config_file)
raise Exception
#do some validation
validate(c)
# get the pipeline name
p_name = c.pipelineName
try:
sublist = yaml.load(open(os.path.realpath(subject_list_file), 'r'))
except:
print("Subject list is not in proper YAML format. Please check your file")
raise Exception
# NOTE: strategies list is only needed in cpac_pipeline prep_workflow for
# creating symlinks
strategies = sorted(build_strategies(c))
print("strategies ---> ")
print(strategies)
sub_scan_map ={}
print("subject list: ")
print(sublist)
try:
for sub in sublist:
if sub['unique_id']:
s = sub['subject_id']+"_" + sub["unique_id"]
else:
s = sub['subject_id']
scan_ids = ['scan_anat']
for id in sub['rest']:
scan_ids.append('scan_'+ str(id))
sub_scan_map[s] = scan_ids
except:
print("\n\n" + "ERROR: Subject list file not in proper format - check if you loaded the correct file?" + "\n" + \
"Error name: cpac_runner_0001" + "\n\n")
raise Exception
create_group_log_template(sub_scan_map, os.path.join(c.outputDirectory, 'logs'))
seeds_created = []
if not (c.seedSpecificationFile is None):
try:
if os.path.exists(c.seedSpecificationFile):
seeds_created = create_seeds_(c.seedOutputLocation, c.seedSpecificationFile, c.FSLDIR)
print('seeds created %s -> ' % seeds_created)
except:
raise IOError('Problem in seedSpecificationFile')
if 1 in c.runVoxelTimeseries:
if 'roi_voxelwise' in c.useSeedInAnalysis:
c.maskSpecificationFile = append_seeds_to_file(c.workingDirectory, seeds_created, c.maskSpecificationFile)
if 1 in c.runROITimeseries:
if 'roi_average' in c.useSeedInAnalysis:
c.roiSpecificationFile = append_seeds_to_file(c.workingDirectory, seeds_created, c.roiSpecificationFile)
if 1 in c.runSCA:
if 'roi_average' in c.useSeedInAnalysis:
c.roiSpecificationFileForSCA = append_seeds_to_file(c.workingDirectory, seeds_created, c.roiSpecificationFileForSCA)
if 1 in c.runNetworkCentrality:
if 'centrality_outputs_smoothed' in c.useSeedInAnalysis:
c.templateSpecificationFile = append_seeds_to_file(c.workingDirectory, seeds_created, c.templateSpecificationFile)
pipeline_timing_info = []
pipeline_timing_info.append(unique_pipeline_id)
pipeline_timing_info.append(pipeline_start_stamp)
pipeline_timing_info.append(len(sublist))
if not c.runOnGrid:
# Import packages
from CPAC.pipeline.cpac_pipeline import prep_workflow
# Init variables
procss = [Process(target=prep_workflow,
args=(sub, c, strategies, 1,
pipeline_timing_info, p_name)) \
for sub in sublist]
pid = open(os.path.join(c.outputDirectory, 'pid.txt'), 'w')
# Init job queue
jobQueue = []
# If we're allocating more processes than are subjects, run them all
if len(sublist) <= c.numSubjectsAtOnce:
"""
Stream all the subjects as sublist is
less than or equal to the number of
subjects that need to run
"""
for p in procss:
p.start()
print(p.pid, file=pid)
# Otherwise manage resources to run processes incrementally
else:
"""
Stream the subject workflows for preprocessing.
At Any time in the pipeline c.numSubjectsAtOnce
will run, unless the number remaining is less than
the value of the parameter stated above
"""
idx = 0
while(idx < len(sublist)):
# If the job queue is empty and we haven't started indexing
if len(jobQueue) == 0 and idx == 0:
# Init subject process index
idc = idx
# Launch processes (one for each subject)
for p in procss[idc: idc + c.numSubjectsAtOnce]:
p.start()
print(p.pid, file=pid)
jobQueue.append(p)
idx += 1
# Otherwise, jobs are running - check them
else:
# Check every job in the queue's status
for job in jobQueue:
# If the job is not alive
if not job.is_alive():
# Find job and delete it from queue
print('found dead job ', job)
loc = jobQueue.index(job)
del jobQueue[loc]
# ...and start the next available process (subject)
procss[idx].start()
# Append this to job queue and increment index
jobQueue.append(procss[idx])
idx += 1
# Add sleep so while loop isn't consuming 100% of CPU
time.sleep(2)
pid.close()
else:
import subprocess
import pickle
temp_files_dir = os.path.join(os.getcwd(), 'cluster_temp_files')
print(subprocess.getoutput("mkdir -p %s" % temp_files_dir))
strategies_file = os.path.join(temp_files_dir, 'strategies.obj')
f = open(strategies_file, 'w')
pickle.dump(strategies, f)
f.close()
if 'sge' in c.resourceManager.lower():
run_sge_jobs(c, config_file, strategies_file, subject_list_file, p_name)
elif 'pbs' in c.resourceManager.lower():
run_pbs_jobs(c, config_file, strategies_file, subject_list_file, p_name)
elif 'condor' in c.resourceManager.lower():
run_condor_jobs(c, config_file, strategies_file, subject_list_file, p_name)
def run(config_file, subject_list_file, p_name = None):
# take date+time stamp for run identification purposes
unique_pipeline_id = strftime("%Y%m%d%H%M%S")
pipeline_start_stamp = strftime("%Y-%m-%d_%H:%M:%S")
try:
if not os.path.exists(config_file):
raise IOError
else:
c = Configuration(yaml.load(open(os.path.realpath(config_file), 'r')))
except IOError:
print "config file %s doesn't exist" % config_file
raise
except Exception:
print "Error reading config file - %s" % config_file
raise Exception
#do some validation
validate(c)
try:
sublist = yaml.load(open(os.path.realpath(subject_list_file), 'r'))
except:
print "Subject list is not in proper YAML format. Please check your file"
raise Exception
strategies = sorted(build_strategies(c))
print "strategies ---> "
print strategies
sub_scan_map ={}
print "subject list: "
print sublist
try:
for sub in sublist:
if sub['unique_id']:
s = sub['subject_id']+"_" + sub["unique_id"]
else:
s = sub['subject_id']
scan_ids = ['scan_anat']
for id in sub['rest']:
scan_ids.append('scan_'+ str(id))
sub_scan_map[s] = scan_ids
except:
print "\n\n" + "ERROR: Subject list file not in proper format - check if you loaded the correct file?" + "\n" + \
"Error name: cpac_runner_0001" + "\n\n"
raise Exception
create_group_log_template(sub_scan_map, os.path.join(c.outputDirectory, 'logs'))
seeds_created = []
if not (c.seedSpecificationFile is None):
try:
if os.path.exists(c.seedSpecificationFile):
seeds_created = create_seeds_(c.seedOutputLocation, c.seedSpecificationFile, c.FSLDIR)
print 'seeds created %s -> ' % seeds_created
except:
raise IOError('Problem in seedSpecificationFile')
if 1 in c.runVoxelTimeseries:
if 'roi_voxelwise' in c.useSeedInAnalysis:
c.maskSpecificationFile = append_seeds_to_file(c.workingDirectory, seeds_created, c.maskSpecificationFile)
if 1 in c.runROITimeseries:
if 'roi_average' in c.useSeedInAnalysis:
c.roiSpecificationFile = append_seeds_to_file(c.workingDirectory, seeds_created, c.roiSpecificationFile)
if 1 in c.runNetworkCentrality:
if 'centrality_outputs_smoothed' in c.useSeedInAnalysis:
c.templateSpecificationFile = append_seeds_to_file(c.workingDirectory, seeds_created, c.templateSpecificationFile)
pipeline_timing_info = []
pipeline_timing_info.append(unique_pipeline_id)
pipeline_timing_info.append(pipeline_start_stamp)
pipeline_timing_info.append(len(sublist))
if not c.runOnGrid:
from CPAC.pipeline.cpac_pipeline import prep_workflow
procss = [Process(target=prep_workflow, args=(sub, c, strategies, 1, pipeline_timing_info, p_name)) for sub in sublist]
pid = open(os.path.join(c.outputDirectory, 'pid.txt'), 'w')
jobQueue = []
if len(sublist) <= c.numSubjectsAtOnce:
"""
Stream all the subjects as sublist is
less than or equal to the number of
subjects that need to run
"""
for p in procss:
p.start()
print >>pid,p.pid
else:
"""
Stream the subject workflows for preprocessing.
At Any time in the pipeline c.numSubjectsAtOnce
will run, unless the number remaining is less than
the value of the parameter stated above
"""
idx = 0
while(idx < len(sublist)):
if len(jobQueue) == 0 and idx == 0:
idc = idx
for p in procss[idc: idc + c.numSubjectsAtOnce]:
p.start()
print >>pid,p.pid
jobQueue.append(p)
idx += 1
else:
for job in jobQueue:
if not job.is_alive():
print 'found dead job ', job
loc = jobQueue.index(job)
del jobQueue[loc]
procss[idx].start()
jobQueue.append(procss[idx])
idx += 1
pid.close()
else:
import commands
import pickle
temp_files_dir = os.path.join(os.getcwd(), 'cluster_temp_files')
print commands.getoutput("mkdir -p %s" % temp_files_dir)
strategies_file = os.path.join(temp_files_dir, 'strategies.obj')
f = open(strategies_file, 'w')
pickle.dump(strategies, f)
f.close()
if 'sge' in c.resourceManager.lower():
run_sge_jobs(c, config_file, strategies_file, subject_list_file, p_name)
elif 'pbs' in c.resourceManager.lower():
run_pbs_jobs(c, config_file, strategies_file, subject_list_file, p_name)
elif 'condor' in c.resourceManager.lower():
run_condor_jobs(c, config_file, strategies_file, subject_list_file, p_name)
def run(config_file, subject_list_file, p_name=None):
# Import packages
import time
# take date+time stamp for run identification purposes
unique_pipeline_id = strftime("%Y%m%d%H%M%S")
pipeline_start_stamp = strftime("%Y-%m-%d_%H:%M:%S")
try:
if not os.path.exists(config_file):
raise IOError
else:
c = Configuration(
yaml.load(open(os.path.realpath(config_file), 'r')))
except IOError:
print "config file %s doesn't exist" % config_file
raise
except Exception:
print "Error reading config file - %s" % config_file
raise Exception
#do some validation
validate(c)
# get the pipeline name
p_name = c.pipelineName
try:
sublist = yaml.load(open(os.path.realpath(subject_list_file), 'r'))
except:
print "Subject list is not in proper YAML format. Please check your file"
raise Exception
# NOTE: strategies list is only needed in cpac_pipeline prep_workflow for
# creating symlinks
strategies = sorted(build_strategies(c))
print "strategies ---> "
print strategies
sub_scan_map = {}
print "subject list: "
print sublist
try:
for sub in sublist:
if sub['unique_id']:
s = sub['subject_id'] + "_" + sub["unique_id"]
else:
s = sub['subject_id']
scan_ids = ['scan_anat']
for id in sub['rest']:
scan_ids.append('scan_' + str(id))
sub_scan_map[s] = scan_ids
except:
print "\n\n" + "ERROR: Subject list file not in proper format - check if you loaded the correct file?" + "\n" + \
"Error name: cpac_runner_0001" + "\n\n"
raise Exception
create_group_log_template(sub_scan_map,
os.path.join(c.outputDirectory, 'logs'))
seeds_created = []
if not (c.seedSpecificationFile is None):
try:
if os.path.exists(c.seedSpecificationFile):
seeds_created = create_seeds_(c.seedOutputLocation,
c.seedSpecificationFile,
c.FSLDIR)
print 'seeds created %s -> ' % seeds_created
except:
raise IOError('Problem in seedSpecificationFile')
if 1 in c.runVoxelTimeseries:
if 'roi_voxelwise' in c.useSeedInAnalysis:
c.maskSpecificationFile = append_seeds_to_file(
c.workingDirectory, seeds_created, c.maskSpecificationFile)
if 1 in c.runROITimeseries:
if 'roi_average' in c.useSeedInAnalysis:
c.roiSpecificationFile = append_seeds_to_file(
c.workingDirectory, seeds_created, c.roiSpecificationFile)
if 1 in c.runSCA:
if 'roi_average' in c.useSeedInAnalysis:
c.roiSpecificationFileForSCA = append_seeds_to_file(
c.workingDirectory, seeds_created,
c.roiSpecificationFileForSCA)
if 1 in c.runNetworkCentrality:
if 'centrality_outputs_smoothed' in c.useSeedInAnalysis:
c.templateSpecificationFile = append_seeds_to_file(
c.workingDirectory, seeds_created, c.templateSpecificationFile)
pipeline_timing_info = []
pipeline_timing_info.append(unique_pipeline_id)
pipeline_timing_info.append(pipeline_start_stamp)
pipeline_timing_info.append(len(sublist))
if not c.runOnGrid:
# Import packages
from CPAC.pipeline.cpac_pipeline import prep_workflow
# Init variables
procss = [Process(target=prep_workflow,
args=(sub, c, strategies, 1,
pipeline_timing_info, p_name)) \
for sub in sublist]
pid = open(os.path.join(c.outputDirectory, 'pid.txt'), 'w')
# Init job queue
jobQueue = []
# If we're allocating more processes than are subjects, run them all
if len(sublist) <= c.numSubjectsAtOnce:
"""
Stream all the subjects as sublist is
less than or equal to the number of
subjects that need to run
"""
for p in procss:
p.start()
print >> pid, p.pid
# Otherwise manage resources to run processes incrementally
else:
"""
Stream the subject workflows for preprocessing.
At Any time in the pipeline c.numSubjectsAtOnce
will run, unless the number remaining is less than
the value of the parameter stated above
"""
idx = 0
while (idx < len(sublist)):
# If the job queue is empty and we haven't started indexing
if len(jobQueue) == 0 and idx == 0:
# Init subject process index
idc = idx
# Launch processes (one for each subject)
for p in procss[idc:idc + c.numSubjectsAtOnce]:
p.start()
print >> pid, p.pid
jobQueue.append(p)
idx += 1
# Otherwise, jobs are running - check them
else:
# Check every job in the queue's status
for job in jobQueue:
# If the job is not alive
if not job.is_alive():
# Find job and delete it from queue
print 'found dead job ', job
loc = jobQueue.index(job)
del jobQueue[loc]
# ...and start the next available process (subject)
procss[idx].start()
# Append this to job queue and increment index
jobQueue.append(procss[idx])
idx += 1
# Add sleep so while loop isn't consuming 100% of CPU
time.sleep(2)
pid.close()
else:
import commands
import pickle
temp_files_dir = os.path.join(os.getcwd(), 'cluster_temp_files')
print commands.getoutput("mkdir -p %s" % temp_files_dir)
strategies_file = os.path.join(temp_files_dir, 'strategies.obj')
f = open(strategies_file, 'w')
pickle.dump(strategies, f)
f.close()
if 'sge' in c.resourceManager.lower():
run_sge_jobs(c, config_file, strategies_file, subject_list_file,
p_name)
elif 'pbs' in c.resourceManager.lower():
run_pbs_jobs(c, config_file, strategies_file, subject_list_file,
p_name)
elif 'condor' in c.resourceManager.lower():
run_condor_jobs(c, config_file, strategies_file, subject_list_file,
p_name)
def run(config_file,
subject_list_file,
p_name=None,
plugin=None,
plugin_args=None):
'''
'''
# Import packages
import commands
import os
import pickle
import time
from CPAC.pipeline.cpac_pipeline import prep_workflow
# Init variables
config_file = os.path.realpath(config_file)
subject_list_file = os.path.realpath(subject_list_file)
# take date+time stamp for run identification purposes
unique_pipeline_id = strftime("%Y%m%d%H%M%S")
pipeline_start_stamp = strftime("%Y-%m-%d_%H:%M:%S")
# Load in pipeline config file
try:
if not os.path.exists(config_file):
raise IOError
else:
c = Configuration(yaml.load(open(config_file, 'r')))
except IOError:
print "config file %s doesn't exist" % config_file
raise
except Exception as e:
raise Exception("Error reading config file - {0}\n\nError details:"
"\n{1}\n\n".format(config_file, e))
c.logDirectory = os.path.abspath(c.logDirectory)
c.workingDirectory = os.path.abspath(c.workingDirectory)
c.outputDirectory = os.path.abspath(c.outputDirectory)
c.crashLogDirectory = os.path.abspath(c.crashLogDirectory)
# Do some validation
validate(c)
# Get the pipeline name
p_name = p_name or c.pipelineName
# Load in subject list
try:
sublist = yaml.load(open(subject_list_file, 'r'))
except:
print "Subject list is not in proper YAML format. Please check " \
"your file"
raise Exception
# NOTE: strategies list is only needed in cpac_pipeline prep_workflow for
# creating symlinks
strategies = sorted(build_strategies(c))
# Populate subject scan map
sub_scan_map = {}
try:
for sub in sublist:
if sub['unique_id']:
s = sub['subject_id'] + "_" + sub["unique_id"]
else:
s = sub['subject_id']
scan_ids = ['scan_anat']
try:
for id in sub['func']:
scan_ids.append('scan_' + str(id))
except KeyError:
for id in sub['rest']:
scan_ids.append('scan_' + str(id))
sub_scan_map[s] = scan_ids
except:
print "\n\n" + "ERROR: Subject list file not in proper format - " \
"check if you loaded the correct file?" + "\n" + \
"Error name: cpac_runner_0001" + "\n\n"
raise Exception
create_group_log_template(sub_scan_map, c.logDirectory)
pipeline_timing_info = []
pipeline_timing_info.append(unique_pipeline_id)
pipeline_timing_info.append(pipeline_start_stamp)
pipeline_timing_info.append(len(sublist))
# If we're running on cluster, execute job scheduler
if c.runOnGrid:
# Create cluster log dir
cluster_files_dir = os.path.join(c.logDirectory, 'cluster_files')
if not os.path.exists(cluster_files_dir):
os.makedirs(cluster_files_dir)
# Create strategies file
strategies_file = os.path.join(cluster_files_dir, 'strategies.obj')
with open(strategies_file, 'w') as f:
pickle.dump(strategies, f)
# Check if its a condor job, and run that
if 'condor' in c.resourceManager.lower():
run_condor_jobs(c, config_file, strategies_file, subject_list_file,
p_name)
# All other schedulers are supported
else:
run_cpac_on_cluster(config_file, subject_list_file,
strategies_file, cluster_files_dir)
# Run on one computer
else:
# Init variables
procss = [
Process(target=prep_workflow,
args=(sub, c, strategies, 1, pipeline_timing_info, p_name,
plugin, plugin_args)) for sub in sublist
]
if not os.path.exists(c.workingDirectory):
try:
os.makedirs(c.workingDirectory)
except:
err = "\n\n[!] CPAC says: Could not create the working " \
"directory: %s\n\nMake sure you have permissions " \
"to write to this directory.\n\n" % c.workingDirectory
raise Exception(err)
pid = open(os.path.join(c.workingDirectory, 'pid.txt'), 'w')
# Init job queue
jobQueue = []
# If we're allocating more processes than are subjects, run them all
if len(sublist) <= c.numParticipantsAtOnce:
for p in procss:
p.start()
print >> pid, p.pid
# Otherwise manage resources to run processes incrementally
else:
idx = 0
while idx < len(sublist):
# If the job queue is empty and we haven't started indexing
if len(jobQueue) == 0 and idx == 0:
# Init subject process index
idc = idx
# Launch processes (one for each subject)
for p in procss[idc:idc + c.numParticipantsAtOnce]:
p.start()
print >> pid, p.pid
jobQueue.append(p)
idx += 1
# Otherwise, jobs are running - check them
else:
# Check every job in the queue's status
for job in jobQueue:
# If the job is not alive
if not job.is_alive():
# Find job and delete it from queue
print 'found dead job ', job
loc = jobQueue.index(job)
del jobQueue[loc]
# ...and start the next available process
# (subject)
procss[idx].start()
# Append this to job queue and increment index
jobQueue.append(procss[idx])
idx += 1
# Add sleep so while loop isn't consuming 100% of CPU
time.sleep(2)
# Close PID txt file to indicate finish
pid.close()
def run(config_file, subject_list_file, p_name=None):
try:
if not os.path.exists(config_file):
raise IOError
else:
c = Configuration(
yaml.load(open(os.path.realpath(config_file), 'r')))
except IOError:
print "config file %s doesn't exist" % config_file
raise
except Exception:
print "Error reading config file - %s" % config_file
raise Exception
#do some validation
validate(c)
try:
sublist = yaml.load(open(os.path.realpath(subject_list_file), 'r'))
except:
print "Subject list is not in proper YAML format. Please check your file"
raise Exception
strategies = sorted(build_strategies(c))
print "strategies ---> "
print strategies
sub_scan_map = {}
print "subject list: "
print sublist
try:
for sub in sublist:
if sub['unique_id']:
s = sub['subject_id'] + "_" + sub["unique_id"]
else:
s = sub['subject_id']
scan_ids = ['scan_anat']
for id in sub['rest']:
scan_ids.append('scan_' + str(id))
sub_scan_map[s] = scan_ids
except:
print "\n\n" + "ERROR: Subject list file not in proper format - check if you loaded the correct file?" + "\n" + \
"Error name: cpac_runner_0001" + "\n\n"
raise Exception
create_group_log_template(sub_scan_map,
os.path.join(c.outputDirectory, 'logs'))
seeds_created = []
if not (c.seedSpecificationFile is None):
try:
if os.path.exists(c.seedSpecificationFile):
seeds_created = create_seeds_(c.seedOutputLocation,
c.seedSpecificationFile,
c.FSLDIR)
print 'seeds created %s -> ' % seeds_created
except:
raise IOError('Problem in seedSpecificationFile')
if 1 in c.runVoxelTimeseries:
if 'roi_voxelwise' in c.useSeedInAnalysis:
c.maskSpecificationFile = append_seeds_to_file(
c.workingDirectory, seeds_created, c.maskSpecificationFile)
if 1 in c.runROITimeseries:
if 'roi_average' in c.useSeedInAnalysis:
c.roiSpecificationFile = append_seeds_to_file(
c.workingDirectory, seeds_created, c.roiSpecificationFile)
if 1 in c.runNetworkCentrality:
if 'centrality_outputs_smoothed' in c.useSeedInAnalysis:
c.templateSpecificationFile = append_seeds_to_file(
c.workingDirectory, seeds_created, c.templateSpecificationFile)
if not c.runOnGrid:
from CPAC.pipeline.cpac_pipeline import prep_workflow
procss = [
Process(target=prep_workflow, args=(sub, c, strategies, 1, p_name))
for sub in sublist
]
pid = open(os.path.join(c.outputDirectory, 'pid.txt'), 'w')
jobQueue = []
if len(sublist) <= c.numSubjectsAtOnce:
"""
Stream all the subjects as sublist is
less than or equal to the number of
subjects that need to run
"""
for p in procss:
p.start()
print >> pid, p.pid
else:
"""
Stream the subject workflows for preprocessing.
At Any time in the pipeline c.numSubjectsAtOnce
will run, unless the number remaining is less than
the value of the parameter stated above
"""
idx = 0
while (idx < len(sublist)):
if len(jobQueue) == 0 and idx == 0:
idc = idx
for p in procss[idc:idc + c.numSubjectsAtOnce]:
p.start()
print >> pid, p.pid
jobQueue.append(p)
idx += 1
else:
for job in jobQueue:
if not job.is_alive():
print 'found dead job ', job
loc = jobQueue.index(job)
del jobQueue[loc]
procss[idx].start()
jobQueue.append(procss[idx])
idx += 1
pid.close()
else:
import commands
import pickle
temp_files_dir = os.path.join(os.getcwd(), 'cluster_temp_files')
print commands.getoutput("mkdir -p %s" % temp_files_dir)
strategies_file = os.path.join(temp_files_dir, 'strategies.obj')
f = open(strategies_file, 'w')
pickle.dump(strategies, f)
f.close()
if 'sge' in c.resourceManager.lower():
run_sge_jobs(c, config_file, strategies_file, subject_list_file,
p_name)
elif 'pbs' in c.resourceManager.lower():
run_pbs_jobs(c, config_file, strategies_file, subject_list_file,
p_name)
elif 'condor' in c.resourceManager.lower():
run_condor_jobs(c, config_file, strategies_file, subject_list_file,
p_name)
