def main():
args = cT.parseCMD()
passwd = cT.Credentials(args.UserName)
# define all data you want to pull from pimc data files IN ORDER.
fileTypes = ['log','estimator','super','bipart_dens','ntWind']
colNums = [
[4,11,12,13,-4],
[0,1,2,3],
[0,1],
[0,1,2] ]
observables = [
['E','Cv1','Cv2','Cv3','N'],
['rho_s/rho','Wx^2','Wy^2','Wz^2'],
['filmDens','bulkDens'],
['W^2','W_{2d}^2','W_{3d}^2'] ]
# grab types of estimators from above. Remove log file.
estimTypes = list(fileTypes)
if 'log' in estimTypes:
estimTypes.pop(estimTypes.index('log'))
# create list of reduced file names for combined data.
fileNames = []
for f in estimTypes:
fileNames.append(str('Reduced'+str(f.capitalize())+'Data.dat'))
print '\nCreating: ',fileNames,'\n'
# if data doesn't exist, pull it from cluster
if not args.pulled:
print 'Pulling data from cluster.'
# create ssh and sftp instances
ssh = paramiko.SSHClient()
ssh.load_host_keys(os.path.expanduser(
os.path.join("~", ".ssh", "known_hosts")))
ssh.connect('bluemoon-user1.uvm.edu', username=args.UserName,
password=passwd)
sftp = ssh.open_sftp()
# move to desired directory on cluster
sftp.chdir(args.targetDir)
# create list of only seed directory names, get rid of other things
allStuffInDir = sftp.listdir()
#for s in seedDirs: # laptop didnt like this method sometimes. wtf?
# if s[:4] != 'seed':
# seedDirs.pop(seedDirs.index(s))
seedDirs = []
for s in range(len(allStuffInDir)):
if 'seed' in allStuffInDir[s]:
seedDirs.append( allStuffInDir[s] )
# pull all requested file types into organized directories
for s in seedDirs:
newName = cT.returnSeedDirName(s)
if os.path.exists("./"+newName):
sys.exit(newName+' already exists.')
os.makedirs("./"+newName)
os.chdir("./"+newName)
sftp.chdir('./'+s+'/OUTPUT/')
moose = sftp.listdir()
for thing in fileTypes:
for m in moose:
if thing in m:
sftp.get('./'+m, './'+m)
print 'pulled ',thing,' files for ',newName
sftp.chdir('../..')
os.chdir('..')
# Rename files to have seed number replace first three numbers of pimcID.
# Optionally delete all seed directories that were pulled from cluster.
cT.renameFilesInDirecs(args.delDir)
# check for repeated pimcIDs -- broken.
#cT.repeatCheck()
# close instances of sftp and ssh
sftp.close()
ssh.close()
# optionally combine all data of the same temperature into one
# much larger array.
if args.Crunch:
print 'Crunching Data'
cT.crunchData(estimTypes,colNums,observables)
# optionally make a trimmed version of the data files that
# makes all arrays the length of the shortest array.
# NOT NECESSARY BUT SOMETIMES USEFUL.
if args.trimData:
print 'Decided to make trimmed data files'
cT.trimData(fileNames)
def main():
args = cT.parseCMD()
passwd = cT.Credentials(args.UserName)
# define all data you want to pull from pimc data files IN ORDER.
fileTypes = ['log', 'estimator', 'super', 'bipart_dens', 'ntWind']
colNums = [[4, 11, 12, 13, -4], [0, 1, 2, 3], [0, 1], [0, 1, 2]]
observables = [['E', 'Cv1', 'Cv2', 'Cv3', 'N'],
['rho_s/rho', 'Wx^2', 'Wy^2', 'Wz^2'],
['filmDens', 'bulkDens'], ['W^2', 'W_{2d}^2', 'W_{3d}^2']]
# grab types of estimators from above. Remove log file.
estimTypes = list(fileTypes)
if 'log' in estimTypes:
estimTypes.pop(estimTypes.index('log'))
# create list of reduced file names for combined data.
fileNames = []
for f in estimTypes:
fileNames.append(str('Reduced' + str(f.capitalize()) + 'Data.dat'))
print '\nCreating: ', fileNames, '\n'
# if data doesn't exist, pull it from cluster
if not args.pulled:
print 'Pulling data from cluster.'
# create ssh and sftp instances
ssh = paramiko.SSHClient()
ssh.load_host_keys(
os.path.expanduser(os.path.join("~", ".ssh", "known_hosts")))
ssh.connect('bluemoon-user1.uvm.edu',
username=args.UserName,
password=passwd)
sftp = ssh.open_sftp()
# move to desired directory on cluster
sftp.chdir(args.targetDir)
# create list of only seed directory names, get rid of other things
allStuffInDir = sftp.listdir()
#for s in seedDirs: # laptop didnt like this method sometimes. wtf?
# if s[:4] != 'seed':
# seedDirs.pop(seedDirs.index(s))
seedDirs = []
for s in range(len(allStuffInDir)):
if 'seed' in allStuffInDir[s]:
seedDirs.append(allStuffInDir[s])
# pull all requested file types into organized directories
for s in seedDirs:
newName = cT.returnSeedDirName(s)
if os.path.exists("./" + newName):
sys.exit(newName + ' already exists.')
os.makedirs("./" + newName)
os.chdir("./" + newName)
sftp.chdir('./' + s + '/OUTPUT/')
moose = sftp.listdir()
for thing in fileTypes:
for m in moose:
if thing in m:
sftp.get('./' + m, './' + m)
print 'pulled ', thing, ' files for ', newName
sftp.chdir('../..')
os.chdir('..')
# Rename files to have seed number replace first three numbers of pimcID.
# Optionally delete all seed directories that were pulled from cluster.
cT.renameFilesInDirecs(args.delDir)
# check for repeated pimcIDs -- broken.
#cT.repeatCheck()
# close instances of sftp and ssh
sftp.close()
ssh.close()
# optionally combine all data of the same temperature into one
# much larger array.
if args.Crunch:
print 'Crunching Data'
cT.crunchData(estimTypes, colNums, observables)
# optionally make a trimmed version of the data files that
# makes all arrays the length of the shortest array.
# NOT NECESSARY BUT SOMETIMES USEFUL.
if args.trimData:
print 'Decided to make trimmed data files'
cT.trimData(fileNames)
def main():
# set number of equilibration steps
equilNum = 1000
# -------------------------------------------------------------------------
# NOTE: NEW USERS WILL NEED TO CHANGE THESE STRINGS!!
# full path to gensubmit and submit file must be supplied as below.
#
# NOTE: This way of doing subFilePath has been replaced by keeping submit
# in the same directory as stateFiles on local machine in the case of
# submissions from equilibrated states. See later in this script!
genSubPath = '/home/max/Documents/Code/PIMC/SCRIPTS/MTG/MTG_CH_gensubmit.py'
#subFilePath = '/home/max/Documents/Code/PIMC/SCRIPTS/submitscripts/submit'
# commands to add directories to path where blitz, boost, pimc sit.
# not sure why, but this must be done every time for paramiko.
expLibs = 'export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/users/m/t/mtgraves/local/lib ; '
expPBSstuff = 'export PATH=/opt/pbs/bin/:$PATH ; '
# -------------------------------------------------------------------------
# parse cmd line
args = cT.parseCMD()
# get username and password for VACC
passwd = cT.Credentials(args.UserName)
# create ssh and sftp instances
ssh = paramiko.SSHClient()
ssh.load_host_keys(os.path.expanduser(
os.path.join("~", ".ssh", "known_hosts")))
ssh.connect('bluemoon-user2.uvm.edu', username=args.UserName, password=passwd)
sftp = ssh.open_sftp()
# Move to desired directory on cluster from home and if it doesn't
# exist, then create it. This is given by the -t cmd line flag.
try:
sftp.chdir(args.targetDir)
except:
sftp.mkdir(args.targetDir)
sftp.chdir(args.targetDir)
# array of seed numbers, between -L and -H flag from cmd line.
seedNums = np.arange(args.lowSeed,args.highSeed+1)
# keep track of directory you are in in the terminal
workingDir = os.getcwd()
# get list of (g)ce-state files in directory supplied from cmdline.
os.chdir('./stateFiles')
stateFileList = sorted(glob.glob('*state*'))
print 'we have ',len(stateFileList),' different stateFiles'
# get list of (g)ce-log files also.
os.chdir('../logFiles')
logFileList = sorted(glob.glob('*log*'))
# -----------------------------------------------------------------
# build list of restart strings from log files
restartStrList = []
for logFile in logFileList:
with open(logFile) as inFile:
for n, line in enumerate(inFile):
if n == 2:
restartStrList += [line[2:]]
# -----------------------------------------------------------------
# build list of worm constants and COM update radii
wormConstList = []
comUpdateList = []
for s in restartStrList:
wormconst = re.findall(r'-C\s\d*\.?\d*(?:e[-+]\d+)?',s)
comUpdateRad = re.findall(r'-D\s\d*\.?\d*(?:e[-+]\d+)?',s)
wormConstList += wormconst
comUpdateList += comUpdateRad
subNum = 0
numToRepeat = 99999 # not currently set to start from multiple states
if numToRepeat*len(stateFileList) < len(seedNums):
sys.exit('Need to up numToRepeat')
os.chdir('..')
for nummm, seedNum in enumerate(seedNums):
# we want to start a few seeds per state.
if nummm != 0 and nummm%numToRepeat == 0:
subNum += 1
print wormConstList[subNum]
print comUpdateList[subNum]
print stateFileList[subNum]
# create seedXXX directory
seedDirName = cT.returnSeedDirName(int(seedNum))
sftp.mkdir(seedDirName)
sftp.chdir('./'+seedDirName)
# create directory structure inside of each seedXXX direc.
sftp.mkdir('out')
sftp.mkdir('OUTPUT')
# change random number seed in submit file -- this defines hacky.
# NOTE: Must have submit script in current working directory.
subFilePath = os.getcwd()
if subFilePath[-1]!='/':
subFilePath +='/'
subFilePath += args.stateFilesDir+'/../submit'
with open(subFilePath) as inFile, open(subFilePath+'_temp', 'w') as outFile:
for n, line in enumerate(inFile):
if n==0:
if '-p 000' not in line:
sys.exit('Include -p 000 in submit script!')
outFile.write( re.sub(r'-p 000', r'-p '+str(seedNum), line))
# run gensubmit to create initial submit script.
command = ('python '+genSubPath+' '+subFilePath+\
'_temp --cluster=bluemoon '+'-m '+args.memRequest)
subprocess.check_call(command, shell=True)
# store submit file name generated from gensubmit
subFile = glob.glob('*submit-pimc.*')
if len(subFile) != 1:
sys.exit('you have more than one submit file here')
subFile = subFile[0]
# optionally set up submit script to call an executable called
# pimcNoSwaps, which must be located on your path on the cluster.
sF2 = subFile[:-4]+'_temp.pbs'
numOccur = 0
stateNum = 0
with open(subFile) as inFile, open(sF2, 'w') as outFile:
for n, line in enumerate(inFile):
match = re.search(r'pimc',line)
if args.noSwaps:
outFile.write( re.sub(r'pimc', r'pimcNoSwaps', line))
else:
outFile.write(line)
if numOccur%2 == 1:
stateNum += 1
if match != None:
numOccur += 1
os.rename(sF2,subFile)
# we check that there is a directory on the cluster called
# stateFiles that contains exactly the (g)ce-state files
# that are in the directory supplied from cmdline.
# This double-storage is set up to make the user think about
# what they are doing.
try:
sftp.chdir('../stateFiles')
stFs = sorted(sftp.listdir())
if stateFileList != stFs:
sys.exit('ERROR: State files in targeted directory on '+
'cluster are different than on your machine.')
sftp.chdir('../'+seedDirName)
except:
sys.exit('ERROR: Must have state files in directory called '+
'stateFiles in the parent directory of your jobs.')
# -----------------------------------------------------------------
# now go through and replace the worm constants.
sF2 = subFile[:-4]+'_temp.pbs'
numOccur = 0
stateNum = 0
os.chdir(workingDir)
with open(subFile) as inFile, open(sF2, 'w') as outFile:
for n, line in enumerate(inFile):
match = re.search(r'-C ',line)
if match != None:
outFile.write( re.sub(r'-C\s\d*\.?\d*(?:e[-+]\d+)?',
wormConstList[subNum], line))
else:
outFile.write(line)
if numOccur%2 == 1:
stateNum += 1
if match != None:
numOccur += 1
os.rename(sF2,subFile)
# -----------------------------------------------------------------
# now go through and replace (or add a new) COM update radii.
sF2 = subFile[:-4]+'_temp.pbs'
numOccur = 0
stateNum = 0
os.chdir(workingDir)
with open(subFile) as inFile, open(sF2, 'w') as outFile:
for n, line in enumerate(inFile):
match = re.search(r'pimc -T ',line)
match2 = re.search(r'pimcNoSwaps -T ',line)
echoMatch = re.search(r'echo "',line)
matchD = re.search(r'-D ',line)
if echoMatch:
outFile.write(line)
else:
if match != None:
if matchD != None:
outFile.write( re.sub(r'-D\s\d*\.?\d*(?:e[-+]\d+)?',
comUpdateList[subNum], line))
else:
outFile.write( re.sub(r'pimc ',
r'pimc '+comUpdateList[subNum]+' ', line))
elif match2 != None:
if matchD != None:
outFile.write( re.sub(r'-D\s\d*\.?\d*(?:e[-+]\d+)?',
comUpdateList[subNum], line))
else:
outFile.write( re.sub(r'pimcNoSwaps ',
r'pimcNoSwaps '+comUpdateList[subNum]+' ', line))
else:
outFile.write(line)
#try:
# outFile.write(line[:-1]+' '+comUpdateList[stateNum])
#except:
# outFile.write(line)
if numOccur%2 == 1:
stateNum += 1
if match != None:
numOccur += 1
os.rename(sF2,subFile)
# -----------------------------------------------------------------
# now go through and replace the -E #### and -p ### part of the
# submitscript with the specified lines.
sF2 = subFile[:-4]+'_temp.pbs'
numOccur = 0
stateNum = 0
clusterCWDpre = '${PBS_O_WORKDIR}/../stateFiles/'
os.chdir(workingDir)
with open(subFile) as inFile, open(sF2, 'w') as outFile:
for n, line in enumerate(inFile):
match = re.search(r'-E\s\d+',line)
if match != None:
outFile.write( re.sub(r'-E\s\d+','-E '+str(equilNum)+
' -s '+clusterCWDpre+stateFileList[subNum],
line))
else:
outFile.write(line)
if numOccur%2 == 1:
stateNum += 1
if match != None:
numOccur += 1
os.rename(sF2,subFile)
# -----------------------------------------------------------------
# copy submit file over to bluemoon
sftp.put('./'+subFile, './'+subFile)
# optionally submit jobs
if args.submitJobs:
# build submit command
submitStuff = 'qsub '+subFile
changeDir = 'cd '+args.targetDir+'/'+seedDirName+' ; '
subComm = changeDir+expLibs+expPBSstuff+submitStuff
# submit the command
stdin, stdout, stderr = ssh.exec_command(subComm)
print 'Output: ',stdout.readlines()
if stderr.readlines() != []:
print 'Error: ',stderr.readlines()
time.sleep(5)
sftp.chdir('..')
sftp.close()
ssh.close()
def main():
# unique tag for the name of the job to be presented in the scheduler.
uTag = 'S05-T0.50'
# set number of equilibration steps
equilNum = 0
# set number of bins to try
binNum = 50000
# -------------------------------------------------------------------------
# commands to add directories to path where blitz, boost, pimc sit.
# not sure why, but this must be done every time for paramiko.
expLibs = 'export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/users/m/t/mtgraves/local/lib ; '
expPBSstuff = 'export PATH=/opt/pbs/bin/:$PATH ; '
# -------------------------------------------------------------------------
# parse cmd line
args = cT.parseCMD()
# get username and password for VACC
passwd = cT.Credentials(args.UserName)
# create ssh and sftp instances
ssh = paramiko.SSHClient()
ssh.load_host_keys(os.path.expanduser(
os.path.join("~", ".ssh", "known_hosts")))
ssh.connect('bluemoon-user2.uvm.edu', username=args.UserName, password=passwd)
sftp = ssh.open_sftp()
# Move to desired directory on cluster.
sftp.chdir(args.targetDir)
# array of seed numbers, between -L and -H flag from cmd line.
seedNums = np.arange(args.lowSeed,args.highSeed+1)
# run through all seeds.
for seedNum in seedNums:
restartStr = ''
# move to seedXXX direc.
seedDirName = cT.returnSeedDirName(int(seedNum))
sftp.chdir('./'+seedDirName)
# if a resubmit file already exists, delete it.
if 'resubmit-pimc.pbs' in sftp.listdir():
sftp.remove('resubmit-pimc.pbs')
# grab log file string from OUTPUT
sftp.chdir('./OUTPUT')
# get names of log files
logFileNames = []
for f in sftp.listdir():
if '-log-' in f:
logFileNames.append(f)
# Naturally sort the log files (account for +/- floats in name).
# NOTE: This works for gce- files with all the same string
# but varying chemical potential. Any other use must be
# thought about first as this hasn't been tested for other uses!!!
logFileNames = natsort.natsorted(logFileNames)
# build array of restart strings from naturally sorted logfiles
restartStrings = []
oneLessLine = False
for logFileName in logFileNames:
with sftp.open(logFileName) as inFile:
for n,line in enumerate(inFile):
# check for whether the restart string in gce-log has been
# moved up one line. We want to write this to the resubmit
# script.
lineUp = False
if n == 1:
if len(line) < 10:
lineUp = True
continue
else:
restartStrings.append(line[2:-1])
if n == 1:
# check if log file 2nd line is blank (some are, some aren't...
# this is caused by bug in converting trestles (xsede) scripts
# over to bluemoon scripts)
if line == '':
oneLessLine = True
continue
else:
restartStrings.append(line[2:-1])
if n == 2:
if not lineUp:
restartStrings.append(line[2:-1])
if not oneLessLine:
restartStrings.append(line[2:-1])
for nr, restartStr in enumerate(restartStrings):
restartStrings[nr]+=' >> ${PBS_O_WORKDIR}/out/pimc-0.out 2>&1'
restartStrings[nr] = re.sub(r'-E\s\d+',r'-E '+str(equilNum),restartStrings[nr])
restartStrings[nr] = re.sub(r'-S\s\d+',r'-S '+str(binNum),restartStrings[nr])
restartStrings[nr] = re.sub(r'-W\s\d+',r'-W '+str(29),restartStrings[nr])
sftp.chdir('..')
# check if no restart string exists, in this case just skip the seed
# (hackyyyy -- needs to start it over in an intelligent manner.)
if restartStrings == []:
print 'seed number ',str(seedNum),' is empty'
sftp.chdir('..')
continue
else:
numStr = 0
reSubName = 'resubmit-pimc.pbs'
# determine name of the original submit script.
for f in sftp.listdir():
if 'submit-pimc' in f:
subFileName = f
# determine if original submit script was from trestles or bluemoon
tresTOblue = False
with sftp.open(subFileName) as inFile:
for n, line in enumerate(inFile):
if n == 6:
if 'uvm104' in line:
tresTOblue = True
# write resubmit script for the case of original submit script being
# from trestles.
if tresTOblue:
with sftp.open(reSubName,'w') as outFile:
# write beginning statements set up with the appropriate torque
# parameters. Also, submit to the tmp directory for faster i/o.
outFile.write('#!/bin/bash\n\
\n#PBS -S /bin/bash\
\n#PBS -l pmem=1gb,pvmem=1gb\
\n#PBS -l nodes=1:ppn=1\
\n#PBS -l walltime=30:00:00\
\n#PBS -N pimc-%s\
\n#PBS -V\
\n#PBS -j oe\
\n#PBS -o out/pimc-${PBS_JOBID}\
\n#PBS -m n\n\
\nmkdir /tmp/${PBS_JOBID}\
\ncd /tmp/${PBS_JOBID}\
\nmkdir OUTPUT\
\ngzip ${PBS_O_WORKDIR}/OUTPUT/*\
\ncp -r ${PBS_O_WORKDIR}/OUTPUT/* ./OUTPUT/\
\ngunzip OUTPUT/*\
\necho "Starting PBS script submit-pimc.pbs at:`date`" \
\necho " host: ${PBS_O_HOST}"\
\necho " node: `cat ${PBS_NODEFILE}`"\
\necho " jobid: ${PBS_JOBID}"\n\n' % uTag)
# write out the submit string appropriate to the job.
with sftp.open(subFileName) as inFile:
for n, line in enumerate(inFile):
if line[:4] == 'pimc':
outFile.write(restartStrings[numStr])
numStr += 1
else:
continue
# write closing statements to bring data back from the node.
outFile.write('\n\ngzip OUTPUT/*\
\ncp OUTPUT/* ${PBS_O_WORKDIR}/OUTPUT\
\ncd ${PBS_O_WORKDIR}\
\ngunzip ${PBS_O_WORKDIR}/OUTPUT\
\nrm -r /tmp/${PBS_JOBID}')
# write resubmit script for the case of original submit script being
# from bluemoon.
else:
with sftp.open(subFileName) as inFile, sftp.open(reSubName,'w') as outFile:
for n, line in enumerate(inFile):
if line[:4] == 'pimc':
outFile.write(restartStrings[numStr])
numStr += 1
elif r'#PBS -N' in line:
outFile.write(line[:-2]+uTag+'\n')
elif r'mkdir OUTPUT' in line:
outFile.write(line)
outFile.write('gzip ${PBS_O_WORKDIR}/OUTPUT/*\n')
outFile.write('cp -r ${PBS_O_WORKDIR}/OUTPUT/* ./OUTPUT/\n')
outFile.write('gunzip OUTPUT/*\n')
else:
outFile.write(line)
print restartStrings[0],'\n'
# my dumb work-around for allowing writing on cluster to finish.
time.sleep(5)
# -----------------------------------------------------------------
# optionally submit jobs
if args.submitJobs:
#if restartStrings == []:
# continue
#else:
# build submit command
submitStuff = 'qsub '+reSubName
changeDir = 'cd '+args.targetDir+'/'+seedDirName+' ; '
subComm = changeDir+expLibs+expPBSstuff+submitStuff
# submit the command
stdin, stdout, stderr = ssh.exec_command(subComm)
print 'Output: ',stdout.readlines()
if stderr.readlines() != []:
print 'Error: ',stderr.readlines()
sftp.chdir('..')