def makeFileList(name) :
if os.path.exists(self.inputFilesListFile(name)) and self.useCachedFileLists() : return
fileNames = eval(self.sampleDict[name].filesCommand)
assert fileNames, "The command '%s' produced an empty list of files"%self.sampleDict[name].filesCommand
tmpDir,localFileName,globalFileName = self.globalToLocal(self.inputFilesListFile(name))
utils.writePickle(localFileName, fileNames)
self.localToGlobal(tmpDir, localFileName, globalFileName)
def makeFileList(name) :
fName = self.inputFilesListFile(name)
if os.path.exists(fName) and self.useCachedFileLists() : return
fileNames = eval(self.sampleDict[name].filesCommand)
assert fileNames, "The command '%s' produced an empty list of files"%self.sampleDict[name].filesCommand
utils.mkdir(os.path.dirname(fName))
utils.writePickle(fName, zip(fileNames,map(nEventsFile, fileNames)))
def writePickle(self) :
def pickleJar(step) :
inter = set(step.varsToPickle()).intersection(set(['nPass','nFail','outputFileName']))
assert not inter, "%s is trying to pickle %s, which %s reserved for use by analysisStep."%(step.name, str(inter), "is" if len(inter)==1 else "are")
return dict([ (item, getattr(step,item)) for item in step.varsToPickle()+['nPass','nFail']] +
[('outputFileName', getattr(step,'outputFileName').replace(self.outputDir, self.globalDir))])
utils.writePickle( self.pickleFileName,
[ [pickleJar(step) for step in self.steps], self.calculablesUsed, self.leavesUsed] )
def __init__(self, options) :
self.__batch = options.batch
self.__resubmit= options.resubmit
self.__loop = options.loop
self.__profile = options.profile
self.__jobId = options.jobId
self.__tag = options.tag
self.__sample = options.sample
self.__site = options.site if options.site!=None else sites.prefix()
self.__tags = options.tags.split(',') if type(options.tags)==str else options.tags
self.__samples = options.samples.split(',') if type(options.samples)==str else options.samples
self.__omit = options.omit.split(',')
self.__nocheck = options.nocheck
self.__quiet = options.quiet
self.__skip = options.skip
self.moveOutputFiles = (self.__jobId is None) or sites.info(site = self.__site, key = "moveOutputFilesBatch")
self.localStem = "%s/%s"%(sites.info(site = self.__site, key = "localOutputDir" ), self.name)
self.globalStem = "%s/%s"%(sites.info(site = self.__site, key = "globalOutputDir"), self.name)
self.sampleDict = samples.SampleHolder()
map(self.sampleDict.update,self.listOfSampleDictionaries())
if self.__tags is True or self.__samples is True :
for conf in self.configurations :
print conf['tag']
if self.__samples is True :
print '\n'.join(' '+sample.weightedName for sample in self.filteredSamples(conf))
sys.exit(0)
if self.__loop:
os.system("mkdir -p %s"%self.localStem)
if self.__jobId==None :
os.system("mkdir -p %s"%self.globalStem)
self.makeInputFileLists()
for conf in self.configurations:
if options.slices < 0:
nSlicesFixed = None
nEventsPerSlice = -options.slices
else:
nSlicesFixed = options.slices
nEventsPerSlice = None
self.listsOfLoopers[conf['tag']] = self.sampleLoopers(conf, nSlicesFixed=nSlicesFixed, nEventsPerSlice=nEventsPerSlice, byEvents=options.byEvents)
if self.__jobId is None and self.__loop:
for looper in self.listsOfLoopers[conf['tag']]:
utils.writePickle(self.jobsFile(conf['tag'], looper.name, clean=True), looper.nSlices)
def __init__(self, options) :
self.__batch = options.batch
self.__loop = int(options.loop) if options.loop!=None else None
self.__nSlices = int(options.slices) if options.slices!=None else 1
self.__profile = options.profile
self.__jobId = options.jobId
self.__tag = options.tag
self.__sample = options.sample
self.__site = options.site if options.site!=None else configuration.sitePrefix()
self.__tags = options.tags.split(',') if type(options.tags)==str else options.tags
self.__samples = options.samples.split(',') if type(options.samples)==str else options.samples
self.__omit = options.omit.split(',')
self.__nocheck = options.nocheck
self.localStem = "%s/%s"%(configuration.siteInfo(site = self.__site, key = "localOutputDir" ), self.name)
self.globalStem = "%s/%s"%(configuration.siteInfo(site = self.__site, key = "globalOutputDir"), self.name)
self.sampleDict = samples.SampleHolder()
map(self.sampleDict.update,self.listOfSampleDictionaries())
if self.__tags is True or self.__samples is True :
for conf in self.configurations :
print conf['tag']
if self.__samples is True :
print '\n'.join(' '+sample.weightedName for sample in self.filteredSamples(conf))
sys.exit(0)
if self.__loop!=None :
os.system("mkdir -p %s"%self.localStem)
if self.__jobId==None :
os.system("mkdir -p %s"%self.globalStem)
self.makeInputFileLists()
for conf in self.configurations :
self.listsOfLoopers[conf['tag']] = self.sampleLoopers(conf)
if self.__jobId==None and self.__loop!=None :
for looper in self.listsOfLoopers[conf['tag']] :
utils.writePickle( self.jobsFile(conf['tag'],looper.name), self.__nSlices )
#t = timeit.Timer(stmt=code,setup='from __main__ import utils')
#print(filename+'_'+str(i+1)+'.pickle')
#print ('%f'%float(t.timeit(10/10)))
sock = getConn()
sock.sendall('token' + filename + '_' + str(i) + '.pickle' + "\n")
received = sock.recv(1024)
print "Received: {}".format(received)
data = open(
'./pickles/token' + filename + '_' + str(i) + '.pickle',
'rb').read()
sock.sendall(data)
sock.close()
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('--filename', '-f', help='Path of file to encrypt')
parser.add_argument('--window', '-w', help='window size of the bucket')
parser.add_argument('--chars', '-c', help='number of chars')
args = parser.parse_args()
#Same key used for different files!
key = nacl.utils.random(nacl.secret.SecretBox.KEY_SIZE)
nonce = nacl.utils.random(nacl.secret.SecretBox.NONCE_SIZE)
utils.writePickle('./pickles/key.pickle', key)
parseRealData(args.filename, int(args.window), int(args.chars))
#createData()
#sendFM()
#createTokens()
def worker(pickle, json) :
if not os.path.exists(pickle) : utils.writePickle(pickle, recordedInvMicrobarns(json))
def train():
X, Y, testX, testFid = prepare_data()
X, Y, testX = str2float(X, Y, testX)
X, testX = scale_data(X, testX)
if args.select_feature:
# select feature
pass
for iter in range(args.sample_num):
np.random.seed(iter)
logger.info('Iteration %2d, Current random seed: %2d' %
(iter, np.random.get_state()[1][0]))
# sampling data
# trainX, trainY, validX, validY = sample_data(X, Y)
trainX = X
trainY = Y
global dTrain
dTrain = xgb.DMatrix(trainX, trainY, nthread=args.nthread)
global dTest
dTest = xgb.DMatrix(testX, nthread=args.nthread)
# bayes_opt selection
# parameter to be learning
logger.info('Setting parameters for BayesianOptimaization')
params = {
'max_depth': (10, args.xgb_max_depth),
'subsample': (args.xgb_subsample, 1),
'min_child_weight': (1, args.xgb_min_child_weight),
'alpha': (0, args.xgb_alpha),
'gamma': (0, args.xgb_gamma),
'colsample_bytree': (args.xgb_colsample_bytree, 1),
'colsample_bylevel': (args.xgb_colsample_bylevel, 1),
'learning_rate':
(args.xgb_learning_rate_lower, args.xgb_learning_rate_upper)
}
logger.info('Running BayesianOptimization')
xgb_bayesopt = BayesianOptimization(train_xgb, params)
xgb_bayesopt.maximize(init_points=5, n_iter=25)
# get the best param
best_params = xgb_bayesopt.res['max']['max_params']
logger.info('Iteration: %d, XGBoost max auc: %f' %
(iter, xgb_bayesopt.res['max']['max_val']))
for param, val in best_params.items():
logger.info('Param %s: %r' % (param, val))
# setting xgboost param
logger.info('Setting best parameters for BayesianOptimization')
xgb_params = {
'nthread': args.nthread,
'n_estimators': args.xgb_n_estimators,
'eta': args.xgb_eta,
'silent': args.xgb_silent,
# for _train_internal
'eval_metric': [args.xgb_eval_metric],
######################
'max_depth': int(best_params['max_depth']),
'subsample': max(min(best_params['subsample'], 1), 0),
'min_child_weight': int(best_params['min_child_weight']),
'alpha': max(best_params['alpha'], 0),
'gamma': max(best_params['gamma'], 0),
'colsample_bytree': max(min(best_params['colsample_bytree'], 1),
0),
'colsample_bylevel': max(min(best_params['colsample_bylevel'], 1),
0),
'learning_rate': max(min(best_params['learning_rate'], 1), 0),
}
# training
model = xgb.train(xgb_params,
dTrain,
num_boost_round=args.xgb_num_boost_rounds,
verbose_eval=args.xgb_verbose_eval,
maximize=args.xgb_maximize)
writePickle(
model,
os.path.join(
'mdl', 'model_iter%d_%dfold_%f.pkl' %
(iter, args.xgb_nfold, xgb_bayesopt.res['max']['max_val'])))
# predict valid y
predY = model.predict(dTest)
result_df = pd.DataFrame(data={'y': predY})
joined_df = pd.DataFrame(testFid).join(result_df)
joined_df.to_csv(os.path.join(
'result', 'xgb_result%d_%dfold.csv' % (iter, args.xgb_nfold)),
index=False)
# re-sorted the fid because of the random splitting data
logger.info(
'----------------------------------------------------------------------\n\n\n'
)
