def __enter__(self):
from cPickle import PickleError
try:
crossValidInfo = load( self._filePath )
except PickleError:
# It failed without renaming the module, retry renaming old module
# structure to new one.
import sys
sys.modules['FastNetTool.CrossValid'] = sys.modules[__name__]
crossValidInfo = load( self._filePath )
# Open crossValidFile:
try:
if isinstance(crossValidInfo, dict):
if crossValidInfo['type'] != 'CrossValidFile':
self._logger.fatal(("Input crossValid file is not from PreProcFile "
"type."))
if crossValidInfo['version'] == 2:
crossValid = CrossValid.fromRawObj( crossValidInfo['crossValid'] )
elif crossValidInfo['version'] == 1:
crossValid = crossValidInfo['crossValid']
from copy import copy
crossValid = crossValid._cnvObj.treatObj( crossValid, copy( crossValid.__dict__ ) )
else:
self._logger.fatal("Unknown job configuration version.")
elif type(crossValidInfo) == list: # Read legacy files
crossValid = crossValidInfo[3]
else:
self._logger.fatal("Invalid CrossValidFile contents.")
except RuntimeError, e:
self._logger.fatal(("Couldn't read cross validation file file '%s': Reason:"
"\n\t %s") % (self._filePath, e))
def __enter__(self):
from cPickle import PickleError
try:
crossValidInfo = load(self._filePath)
except PickleError:
# It failed without renaming the module, retry renaming old module
# structure to new one.
import sys
sys.modules['FastNetTool.CrossValid'] = sys.modules[__name__]
crossValidInfo = load(self._filePath)
# Open crossValidFile:
try:
if isinstance(crossValidInfo, dict):
if crossValidInfo['type'] != 'CrossValidFile':
self._fatal(
("Input crossValid file is not from PreProcFile "
"type."))
if crossValidInfo['version'] == 2:
crossValid = CrossValid.fromRawObj(
crossValidInfo['crossValid'])
elif crossValidInfo['version'] == 1:
crossValid = crossValidInfo['crossValid']
from copy import copy
crossValid = crossValid._cnvObj.treatObj(
crossValid, copy(crossValid.__dict__))
else:
self._fatal("Unknown job configuration version.")
elif type(crossValidInfo) == list: # Read legacy files
crossValid = crossValidInfo[3]
else:
self._fatal("Invalid CrossValidFile contents.")
except RuntimeError, e:
self._fatal(
("Couldn't read cross validation file file '%s': Reason:"
"\n\t %s") % (self._filePath, e))
def __enter__(self):
# Open file:
jobConfig = load(self._filePath)
try:
if type(jobConfig) is dict:
if jobConfig['type'] != self._type:
self._logger.fatal(("Input jobConfig file is not from jobConfig "
"type."))
# Read configuration file to retrieve pre-processing,
if jobConfig['version'] == 1:
neuronBounds = MatlabLoopingBounds( jobConfig['neuronBounds'] )
sortBounds = PythonLoopingBounds( jobConfig['sortBounds'] )
initBounds = PythonLoopingBounds( jobConfig['initBounds'] )
else:
self._logger.fatal("Unknown job configuration version")
elif type(jobConfig) is list: # zero version file (without versioning
# control):
neuronBounds = MatlabLoopingBounds( [jobConfig[0], jobConfig[0]] )
sortBounds = MatlabLoopingBounds( jobConfig[1] )
initBounds = MatlabLoopingBounds( jobConfig[2] )
else:
self._logger.fatal("Unknown file type entered for config file.")
except RuntimeError, e:
self._logger.fatal(("Couldn't read configuration file '%s': Reason:"
"\n\t %s" % (self._filePath, e)))
def __enter__(self):
# Open file:
jobConfig = load(self._filePath)
try:
if type(jobConfig) is dict:
if jobConfig['type'] != self._type:
self._fatal(("Input jobConfig file is not from jobConfig "
"type."))
# Read configuration file to retrieve pre-processing,
if jobConfig['version'] == 1:
neuronBounds = MatlabLoopingBounds( jobConfig['neuronBounds'] )
sortBounds = PythonLoopingBounds( jobConfig['sortBounds'] )
initBounds = PythonLoopingBounds( jobConfig['initBounds'] )
else:
self._fatal("Unknown job configuration version")
elif type(jobConfig) is list: # zero version file (without versioning
# control):
neuronBounds = MatlabLoopingBounds( [jobConfig[0], jobConfig[0]] )
sortBounds = MatlabLoopingBounds( jobConfig[1] )
initBounds = MatlabLoopingBounds( jobConfig[2] )
else:
self._fatal("Unknown file type entered for config file.")
except RuntimeError, e:
self._fatal(("Couldn't read configuration file '%s': Reason:"
"\n\t %s" % (self._filePath, e)))
def __init__(self, crossvalFileName, monFileName, **kw):
from ROOT import TFile, gROOT
gROOT.ProcessLine("gErrorIgnoreLevel = kFatal;");
#Set all global setting from ROOT style plot!
SetTuningStyle()
Logger.__init__(self, kw)
#Hold all information abount the monitoring root file
self._infoObjs = list()
try:#Protection
self._logger.info('Reading monRootFile (%s)',monFileName)
self._rootObj = TFile(monFileName, 'read')
except RuntimeError:
self._logger.fatal('Could not open root monitoring file.')
from RingerCore import load
try:#Protection
self._logger.info('Reading crossvalFile (%s)',crossvalFileName)
crossvalObj = load(crossvalFileName)
except RuntimeError:
self._logger.fatal('Could not open pickle summary file.')
#Loop over benchmarks
for benchmarkName in crossvalObj.keys():
#Must skip if ppchain collector
if benchmarkName == 'infoPPChain': continue
#Add summary information into MonTuningInfo helper class
self._infoObjs.append( TuningMonitoringInfo( benchmarkName, crossvalObj[benchmarkName] ) )
self._logger.info('Creating MonTuningInfo for %s and the iterator object [et=%d, eta=%d]',
benchmarkName,self._infoObjs[-1].etbin(), self._infoObjs[-1].etabin())
#Loop over all benchmarks
#Always be the same bin for all infoObjs
etabin = self._infoObjs[0].etabin()
etbin = self._infoObjs[0].etbin()
#Reading the data rings from path or object
refFile = kw.pop('refFile', None)
if refFile:
from TuningTools import TuningDataArchieve
TDArchieve = TuningDataArchieve(refFile)
self._logger.info(('Reading perf file with name %s')%(refFile))
try:
with TDArchieve as data:
self._data = (data['signal_patterns'][etbin][etabin], data['background_patterns'][etbin][etabin])
except RuntimeError:
self._logger.fatal('Could not open the patterns data file.')
raise RuntimeError('Could not open the patterns data file.')
else:
patterns = kw.pop('patterns', None)
if patterns:
self._data = (patterns['signal_patterns'][etbin][etabin], patterns['background_patterns'][etbin][etabin])
else:
raise RuntimeError('You must pass the ref file as parameter. abort!')
def __init__(self, crossvalFileName, monFileName, **kw):
from ROOT import TFile, gROOT
gROOT.ProcessLine("gErrorIgnoreLevel = kFatal;")
#Set all global setting from ROOT style plot!
SetTuningStyle()
Logger.__init__(self, kw)
#Hold all information abount the monitoring root file
self._infoObjs = list()
try: #Protection
self._logger.info('Reading monRootFile (%s)', monFileName)
self._rootObj = TFile(monFileName, 'read')
except RuntimeError:
self._logger.fatal('Could not open root monitoring file.')
from RingerCore import load
try: #Protection
self._logger.info('Reading crossvalFile (%s)', crossvalFileName)
crossvalObj = load(crossvalFileName)
except RuntimeError:
self._logger.fatal('Could not open pickle summary file.')
#Loop over benchmarks
for benchmarkName in crossvalObj.keys():
#Must skip if ppchain collector
if benchmarkName == 'infoPPChain': continue
#Add summary information into MonTuningInfo helper class
self._infoObjs.append(
TuningMonitoringInfo(benchmarkName,
crossvalObj[benchmarkName]))
self._logger.info(
'Creating MonTuningInfo for %s and the iterator object [et=%d, eta=%d]',
benchmarkName, self._infoObjs[-1].etbinidx(),
self._infoObjs[-1].etabinidx())
#Loop over all benchmarks
#Always be the same bin for all infoObjs
etabinidx = self._infoObjs[0].etabinidx()
etbinidx = self._infoObjs[0].etbinidx()
#Reading the data rings from path or object
dataPath = kw.pop('dataPath', None)
if dataPath:
from TuningTools import TuningDataArchieve
self._logger.info(
('Reading data tuning file with name %s') % (dataPath))
TDArchieve = TuningDataArchieve.load(dataPath,
etBinIdx=etbinidx,
etaBinIdx=etabinidx,
loadEfficiencies=False)
self._data = (TDArchieve.signalPatterns,
TDArchieve.backgroundPatterns)
else:
self._logger.fatal(
'You must pass the ref file as parameter. abort!')
def __enter__(self):
subsetColInfo = load( self._filePath )
# Open crossValidPreprocFile:
try:
if isinstance(subsetColInfo, dict):
if subsetColInfo['type'] != 'SubsetGeneratorCollectionFile':
self._fatal(("Input subsetCol file is not from SubsetGeneratorCollectionFile type."))
if subsetColInfo['version'] == 0:
self._subsetCol = SubsetGeneratorCollection.fromRawObj( subsetColInfo['subsetCol'] )
else:
self._fatal("Unknown job configuration version.")
else:
self._fatal("Invalid SubsetGeneratorCollectionFile contents.")
except RuntimeError, e:
self._fatal(("Couldn't read cross preproc validation file file '%s': Reason:"
"\n\t %s") % (self._filePath, e))
def __enter__(self):
# Open file:
jobConfig = load(self._filePath)
try:
if type(jobConfig) is dict:
if jobConfig['type'] != self._type:
self._fatal(("Input jobConfig file is not from jobConfig "
"type."))
# Read configuration file to retrieve pre-processing,
if jobConfig['version'] >= 1:
neuronBounds = MatlabLoopingBounds(
jobConfig['neuronBounds'])
sortBounds = PythonLoopingBounds(jobConfig['sortBounds'])
initBounds = PythonLoopingBounds(jobConfig['initBounds'])
modelBounds = [None for _ in neuronBounds()]
if jobConfig['version'] == 2:
modelBounds = jobConfig['modelBounds']
if modelBounds and coreConf() is TuningToolCores.keras:
from keras.models import model_from_json
import json
for idx, model in enumerate(modelBounds):
modelBounds[idx] = model_from_json(
json.dumps(model, separators=(
',', ':'))) if model else None
else:
self._fatal("Unknown job configuration version")
elif type(jobConfig
) is list: # zero version file (without versioning
# control):
neuronBounds = MatlabLoopingBounds(
[jobConfig[0], jobConfig[0]])
sortBounds = MatlabLoopingBounds(jobConfig[1])
initBounds = MatlabLoopingBounds(jobConfig[2])
modelBounds = None
else:
self._fatal("Unknown file type entered for config file.")
except RuntimeError, e:
self._fatal(("Couldn't read configuration file '%s': Reason:"
"\n\t %s" % (self._filePath, e)))
def changeOp(d, key, mat):
def changePassed(key, mat):
for lkey in d[key].item():
for rowIdx, array1 in enumerate( d[key].item()[lkey]):
for columnIdx, rawdict in enumerate(array1):
rawdict['_passed'] = mat[rowIdx][columnIdx]
def changeTotal(key, mat):
for lkey in d[key].item():
for rowIdx, array1 in enumerate( d[key].item()[lkey]):
for columnIdx, rawdict in enumerate(array1):
rawdict['_count'] = 100.
changePassed( key, mat )
changeTotal( key, mat )
from RingerCore import load, save
a = load('mc15_13TeV.361106.423300.sgn.trigegprobes.bkg.vetotruth.trig.l2calo.eg.std.grid.medium.npz')
d = dict(a)
changeOp( d, 'signal_efficiencies', mergeEffTable( veryloose20160701 ) )
changeOp( d, 'background_efficiencies', pfrefs_vloose )
save(d, 'mc15_13TeV.361106.423300.sgn.trigegprobes.bkg.vetotruth.trig.l2calo.eg.std.grid.veryloose', protocol = 'savez_compressed')
d = dict(a)
changeOp( d, 'signal_efficiencies', mergeEffTable( loose20160701 ) )
changeOp( d, 'background_efficiencies', pfrefs_loose )
save(d, 'mc15_13TeV.361106.423300.sgn.trigegprobes.bkg.vetotruth.trig.l2calo.eg.std.grid.loose', protocol = 'savez_compressed')
d = dict(a)
changeOp( d, 'signal_efficiencies', mergeEffTable( tight20160701 ) )
changeOp( d, 'background_efficiencies', pfrefs_tight )
save(d, 'mc15_13TeV.361106.423300.sgn.trigegprobes.bkg.vetotruth.trig.l2calo.eg.std.grid.tight', protocol = 'savez_compressed')
parser.add_argument('-d',
'--data',
action='store',
dest='data',
required=True,
nargs='+',
help="The input tuning files.")
import sys, os
if len(sys.argv) == 1:
parser.print_help()
sys.exit(1)
args = parser.parse_args()
# Take all files
paths = csvStr2List(args.data)
paths = expandFolders(paths)
from RingerCore import load, save, appendToFileName
for f in paths:
ff = load(f)
for k in ff.keys():
if 'SP' in k:
etBin = ff[k]['etBinIdx']
etaBin = ff[k]['etaBinIdx']
print 'etBin = ', etBin, ', etaBin = ', etaBin
outname = f.split('/')[len(f.split('/')) - 2]
cOutputName = appendToFileName(outname, ('et%d_eta%d') % (etBin, etaBin))
save(ff, cOutputName, compress=True)
npCurrent, _ = retrieve_npConstants()
npCurrent.level = args.output_level
logger = Logger.getModuleLogger(__name__, args.output_level)
files = expandFolders(args.inputs)
from zipfile import BadZipfile
from copy import deepcopy
for f in files:
logger.info("Turning numpy matrix file '%s' into pre-processing file...", f)
fileparts = f.split("/")
folder = "/".join(fileparts[0:-1]) + "/"
fname = fileparts[-1]
try:
data = dict(load(f))
except BadZipfile, e:
logger.warning("Couldn't load file '%s'. Reason:\n%s", f, str(e))
continue
logger.debug("Finished loading file '%s'...", f)
for key in data:
if key == "W":
ppCol = deepcopy(data["W"])
from TuningTools.PreProc import *
for obj, idx, parent, _, _ in traverse(ppCol, tree_types=(np.ndarray,), max_depth=3):
parent[idx] = PreProcChain(RemoveMean(), Projection(matrix=obj), UnitaryRMS())
# Turn arrays into mutable objects:
ppCol = ppCol.tolist()
ppCol = fixPPCol(ppCol, len(ppCol[0][0]), len(ppCol[0]), len(ppCol))
if fname.endswith(".npz"):
def makeSummaryMonSlides(outputs, nbins, choicesfile, grid=False):
from scipy.io import loadmat
import os
f = loadmat(choicesfile)
choices = dict()
choices['Pd'] = f['choices']['Pd'][0][0]
choices['Pf'] = f['choices']['Pf'][0][0]
choices['SP'] = f['choices']['SP'][0][0]
nbinchoices = 0
for etchoices in choices['Pd']:
nbinchoices += len(etchoices)
if not nbins == nbinchoices:
raise NameError('Choices Archieve Error')
net = len(choices['Pd'])
neta = len(choices['Pd'][0])
unbinned = False
if net == 1 and neta == 1:
unbinned = True
if grid:
f = load([s for s in os.listdir('.') if '_et0_eta0' in s][0] +
'/perfBounds.pic.gz')
else:
f = load('{}/perfBounds.pic.gz'.format(outputs + '_et0_eta0'))
benchmarkNames = f['perf'].keys()
for benchmarkName in benchmarkNames:
slideAnex = []
for et in xrange(net):
etlist = []
for eta in xrange(neta):
etaDict = dict()
neuron = choices[benchmarkName.split('_')[-1]][et][eta]
basepath = outputs
basepath += ('_et%d_eta%d') % (et, eta)
if grid:
basepath = ('_et%d_eta%d') % (et, eta)
basepath = [s for s in os.listdir('.') if basepath in s][0]
f = load(basepath + '/perfBounds.pic.gz')
else:
f = load('{}/perfBounds.pic.gz'.format(basepath))
etstr = f['bounds']['etbinstr']
etastr = f['bounds']['etabinstr']
perfs = f['perf'][benchmarkName]
refBench = perfs['config_' + str(neuron).zfill(3)].getRef()
detR = r'{:.2f}'.format(refBench['det'])
spR = r'{:.2f}'.format(refBench['sp'])
faR = r'{:.2f}'.format(refBench['fa'])
refBench = [detR, spR, faR]
perfBench = perfs['config_' + str(neuron).zfill(3)].getPerf()
detP = r'{:.2f}$\pm${:.2f}'.format(perfBench['detMean'],
perfBench['detStd'])
spP = r'{:.2f}$\pm${:.2f}'.format(perfBench['spMean'],
perfBench['spStd'])
faP = r'{:.2f}$\pm${:.2f}'.format(perfBench['faMean'],
perfBench['faStd'])
perfBench = [detP, spP, faP]
bestNetBench = perfs['config_' + str(neuron).zfill(3)].rawOp()
detB = r'{:.2f}'.format(bestNetBench['det'] * 100)
spB = r'{:.2f}'.format(bestNetBench['sp'] * 100)
faB = r'{:.2f}'.format(bestNetBench['fa'] * 100)
bestNetBench = [detB, spB, faB]
perfs = [refBench, perfBench, bestNetBench]
graphSections = [
'All Sorts(Validation)',
'All ROC Sorts(Validation)',
'All Sorts(Operation)',
'All ROC Sorts(Operation)',
'Best Network',
'Best Operation Output',
]
figures = [
'{}/figures/{}/neuron_{}/plot_{}_neuron_{}_sorts_val.pdf'.
format(basepath, benchmarkName, neuron, benchmarkName,
neuron),
'{}/figures/{}/neuron_{}/plot_{}_neuron_{}_sorts_roc_tst.pdf'
.format(basepath, benchmarkName, neuron, benchmarkName,
neuron),
'{}/figures/{}/neuron_{}/plot_{}_neuron_{}_sorts_op.pdf'.
format(basepath, benchmarkName, neuron, benchmarkName,
neuron),
'{}/figures/{}/neuron_{}/plot_{}_neuron_{}_sorts_roc_op.pdf'
.format(basepath, benchmarkName, neuron, benchmarkName,
neuron),
'{}/figures/{}/neuron_{}/plot_{}_neuron_{}_best_op.pdf'.
format(basepath, benchmarkName, neuron, benchmarkName,
neuron),
'{}/figures/{}/neuron_{}/plot_{}_neuron_{}_best_op_output.pdf'
.format(basepath, benchmarkName, neuron, benchmarkName,
neuron),
]
figuresDict = dict(zip(graphSections, figures))
etaDict['neuron'] = neuron
etaDict['figures'] = figuresDict
etaDict['graphSections'] = graphSections
etaDict['perfs'] = perfs
etaDict['etastr'] = etastr
etaDict['etstr'] = etstr
etlist.append(etaDict)
# for eta
slideAnex.append(etlist)
#for et
with BeamerTexReportTemplate1(theme='Berlin',
_toPDF=True,
title=benchmarkName,
outputFile=benchmarkName,
font='structurebold'):
with BeamerSection(name='Performance'):
if not unbinned:
l1 = ['', '']
l2 = ['', '']
sideline = '{|c|c|'
for et in xrange(net):
sideline += 'ccc|'
l1.extend(['', slideAnex[et][0]['etstr'], ''])
l2.extend(['Pd', 'SP', 'PF'])
sideline += '}'
etlines = []
for eta in xrange(neta):
la = [
r'\hline' + slideAnex[0][eta]['etastr'],
'CrossValidation'
]
lb = [r'', 'Reference']
lc = [r'', 'bestNetBench']
for et in xrange(net):
prfs = slideAnex[et][eta]['perfs']
refBench = prfs[0]
perfBench = prfs[1]
bestNetBench = prfs[2]
la.extend(perfBench)
lb.extend(refBench)
lc.extend(bestNetBench)
etlines.extend([la, lb, lc])
linhas = []
linhas.append(l1)
linhas.append(l2)
linhas.extend(etlines)
with BeamerSubSection(name='All Bins Performance'):
BeamerTableSlide(
title='All Bins Performance',
linhas=linhas,
sideline=sideline,
caption='Efficiences',
)
for et in xrange(net):
l1 = ['', '']
l2 = ['', '']
sideline = '{|c|c|ccc|}'
l1.extend(['', slideAnex[et][0]['etstr'], ''])
l2.extend(['Pd', 'SP', 'PF'])
etlines = []
for eta in xrange(neta):
if unbinned:
la = [r'\hline', 'CrossValidation']
else:
la = [
r'\hline' + slideAnex[0][eta]['etastr'],
'CrossValidation'
]
lb = [r'Neuron', 'Reference']
lc = [r'', 'bestNetBench']
neuron = slideAnex[et][eta]['neuron']
prfs = slideAnex[et][eta]['perfs']
lc[0] = '%d' % neuron
refBench = prfs[0]
perfBench = prfs[1]
bestNetBench = prfs[2]
la.extend(perfBench)
lb.extend(refBench)
lc.extend(bestNetBench)
etlines.extend([la, lb, lc])
linhas = []
if not unbinned:
linhas.append(l1)
linhas.append(l2)
linhas.extend(etlines)
if unbinned:
BeamerTableSlide(
title='Performance',
linhas=linhas,
sideline=sideline,
caption='Efficiences',
)
else:
with BeamerSubSection(name=slideAnex[et][0]['etstr']):
BeamerTableSlide(
title='Performance',
linhas=linhas,
sideline=sideline,
caption='Efficiences',
)
with BeamerSection(name='Figures'):
if unbinned:
neuron = slideAnex[0][0]['neuron']
graphSections = slideAnex[0][0]['graphSections']
figures = slideAnex[0][0]['figures']
for graph in graphSections:
BeamerFigureSlide(title=graph + ' Neuron: ' +
str(neuron),
path=figures[graph])
else:
for et in xrange(net):
with BeamerSubSection(name=slideAnex[et][0]['etstr']):
for eta in xrange(neta):
neuron = slideAnex[et][eta]['neuron']
graphSections = slideAnex[et][eta][
'graphSections']
figures = slideAnex[et][eta]['figures']
with BeamerSubSubSection(
name=slideAnex[et][eta]['etastr'] +
', Neuron: {}'.format(neuron)):
for graph in graphSections:
BeamerFigureSlide(title=graph,
path=figures[graph])
return
npCurrent.level = args.output_level
logger = Logger.getModuleLogger(__name__, args.output_level)
files = expandFolders(args.inputs)
from zipfile import BadZipfile
from copy import deepcopy
for f in files:
logger.info("Turning numpy matrix file '%s' into pre-processing file...",
f)
fileparts = f.split('/')
folder = '/'.join(fileparts[0:-1]) + '/'
fname = fileparts[-1]
try:
data = dict(load(f))
except BadZipfile, e:
logger.warning("Couldn't load file '%s'. Reason:\n%s", f, str(e))
continue
logger.debug("Finished loading file '%s'...", f)
for key in data:
if key == 'W':
ppCol = deepcopy(data['W'])
from TuningTools.PreProc import *
for obj, idx, parent, _, _ in traverse(ppCol,
tree_types=(np.ndarray, ),
max_depth=3):
parent[idx] = PreProcChain(RemoveMean(),
Projection(matrix=obj),
UnitaryRMS())
# Turn arrays into mutable objects:
args.inputFiles = expandFolders(args.inputFiles)
mainLogger.verbose("All input files are:")
if mainLogger.isEnabledFor(LoggingLevel.VERBOSE):
pprint(args.inputFiles)
for inFile in progressbar(args.inputFiles,
len(args.inputFiles),
logger=mainLogger,
prefix="Processing files "):
# Treat output file name:
from RingerCore import checkExtension, changeExtension, load, save
if checkExtension(inFile, "tgz|tar.gz|pic"):
cOutputName = changeExtension(inFile, '.mat')
if args.change_output_folder:
import os.path
cOutputName = os.path.join(
os.path.abspath(args.change_output_folder),
os.path.basename(cOutputName))
data = load(inFile, useHighLevelObj=False)
from scipy.io import savemat
try:
savemat(cOutputName, data)
except ImportError:
self._logger.fatal(
("Cannot save matlab file, it seems that scipy is not "
"available."), ImportError)
mainLogger.info("Successfully created matlab file: %s", cOutputName)
else:
mainLogger.error("Cannot transform files '%s' to matlab." % inFile)
# end of (for fileCollection)
def getModels(self, summaryInfoList, **kw):
refBenchCol = kw.pop( 'refBenchCol', None )
configCol = kw.pop( 'configCol', [] )
muBin = kw.pop( 'muBin', [-999,9999] )
checkForUnusedVars( kw, self._logger.warning )
# Treat the summaryInfoList
if not isinstance( summaryInfoList, (list,tuple)):
summaryInfoList = [ summaryInfoList ]
summaryInfoList = list(traverse(summaryInfoList,simple_ret=True))
nSummaries = len(summaryInfoList)
if not nSummaries:
logger.fatal("Summary dictionaries must be specified!")
if refBenchCol is None:
refBenchCol = summaryInfoList[0].keys()
# Treat the reference benchmark list
if not isinstance( refBenchCol, (list,tuple)):
refBenchCol = [ refBenchCol ] * nSummaries
if len(refBenchCol) == 1:
refBenchCol = refBenchCol * nSummaries
nRefs = len(list(traverse(refBenchCol,simple_ret=True)))
# Make sure that the lists are the same size as the reference benchmark:
nConfigs = len(list(traverse(configCol,simple_ret=True)))
if nConfigs == 0:
configCol = [None for i in range(nRefs)]
elif nConfigs == 1:
configCol = configCol * nSummaries
nConfigs = len(list(traverse(configCol,simple_ret=True)))
if nConfigs != nRefs:
logger.fatal("Summary size is not equal to the configuration list.", ValueError)
if nRefs == nConfigs == nSummaries:
# If user input data without using list on the configuration, put it as a list:
for o, idx, parent, _, _ in traverse(configCol):
parent[idx] = [o]
for o, idx, parent, _, _ in traverse(refBenchCol):
parent[idx] = [o]
configCol = list(traverse(configCol,max_depth_dist=1,simple_ret=True))
refBenchCol = list(traverse(refBenchCol,max_depth_dist=1,simple_ret=True))
nConfigs = len(configCol)
nSummary = len(refBenchCol)
if nRefs != nConfigs != nSummary:
logger.fatal("Number of references, configurations and summaries do not match!", ValueError)
discrList = []
from itertools import izip, count
for summaryInfo, refBenchmarkList, configList in zip(summaryInfoList,refBenchCol,configCol):
if type(summaryInfo) is str:
self._logger.info('Loading file "%s"...', summaryInfo)
summaryInfo = load(summaryInfo)
elif type(summaryInfo) is dict:
pass
else:
logger.fatal("Cross-valid summary info is not string and not a dictionary.", ValueError)
for idx, refBenchmarkName, config in izip(count(), refBenchmarkList, configList):
try:
key = filter(lambda x: refBenchmarkName in x, summaryInfo)[0]
refDict = summaryInfo[ key ]
except IndexError :
self._logger.fatal("Could not find reference %s in summaryInfo. Available options are: %r", refBenchmarkName, summaryInfo.keys())
self._logger.info("Using Reference key: %s", key )
ppInfo = summaryInfo['infoPPChain']
etBinIdx = refDict['etBinIdx']
etaBinIdx = refDict['etaBinIdx']
etBin = refDict['etBin']
etaBin = refDict['etaBin']
info = refDict['infoOpBest'] if config is None else refDict['config_' + str(config).zfill(3)]['infoOpBest']
# Check if user specified parameters for exporting discriminator
# operation information:
sort = info['sort']
init = info['init']
pyThres = info['cut']
from RingerCore import retrieveRawDict
if isinstance( pyThres, float ):
pyThres = RawThreshold( thres = pyThres
, etBinIdx = etBinIdx, etaBinIdx = etaBinIdx
, etBin = etBin, etaBin = etaBin)
else:
# Get the object from the raw dict
pyThres = retrieveRawDict( pyThres )
if pyThres.etBin is None:
pyThres.etBin = etBin
elif pyThres.etBin is '':
pyThres.etBin = etBin
elif isinstance( pyThres.etBin, (list,tuple)):
pyThres.etBin = np.array( pyThres.etBin)
if not(np.array_equal( pyThres.etBin, etBin )):
self._logger.fatal("etBin does not match for threshold! Should be %r, is %r", pyThres.etBin, etBin )
if pyThres.etaBin is None:
pyThres.etaBin = etaBin
elif pyThres.etaBin is '':
pyThres.etaBin = etaBin
elif isinstance( pyThres.etaBin, (list,tuple)):
pyThres.etaBin = np.array( pyThres.etaBin)
if not(np.array_equal( pyThres.etaBin, etaBin )):
self._logger.fatal("etaBin does not match for threshold! Should be %r, is %r", pyThres.etaBin, etaBin )
if type(pyThres) is RawThreshold:
thresValues = [pyThres.thres]
else:
thresValues = [pyThres.slope, pyThres.intercept, pyThres.rawThres]
pyPreProc = ppInfo['sort_'+str(sort).zfill(3)]['items'][0]
pyPreProc = retrieveRawDict( pyPreProc )
useCaloRings=False; useTrack=False; useShowerShape=False
if type(pyPreProc) is Norm1:
useCaloRings=True
elif type(pyPreProc) is TrackSimpleNorm:
useTrack=True
elif type(pyPreProc) is ShowerShapesSimpleNorm:
useShowerShape=True
elif type(pyPreProc) is ExpertNetworksSimpleNorm:
useCaloRings=True; useTrack=True
elif type(pyPreProc) is ExpertNetworksShowerShapeSimpleNorm:
useCaloRings=True; useShowerShape=True
elif type(pyPreProc) is ExpertNetworksShowerShapeAndTrackSimpleNorm:
useCaloRings=True; useTrack=True; useShowerShape=True
elif type(pyPreProc) is PreProcMerge:
for slot in pyPreProc.slots:
if type(pyPreProc) is Norm1:
useCaloRings=True
elif type(pyPreProc) is TrackSimpleNorm:
useTrack=True
elif type(pyPreProc) is ShowerShapesSimpleNorm:
useShowerShape=True
elif type(pyPreProc) is ExpertNetworksSimpleNorm:
useCaloRings=True; useTrack=True
elif type(pyPreProc) is ExpertNetworksShowerShapeSimpleNorm:
useCaloRings=True; useShowerShape=True
elif type(pyPreProc) is ExpertNetworksShowerShapeAndTrackSimpleNorm:
useCaloRings=True; useTrack=True; useShowerShape=True
else:
self._logger.fatal('PrepProc strategy not found...')
discrDict = info['discriminator']
model = {
'discriminator' : discrDict,
'threshold' : thresValues,
'etBin' : etBin,
'etaBin' : etaBin,
'muBin' : muBin,
'etBinIdx' : etBinIdx,
'etaBinIdx' : etaBinIdx,
}
removeOutputTansigTF = refDict.get('removeOutputTansigTF', None )
model['removeOutputTansigTF'] = removeOutputTansigTF
model['useCaloRings'] = useCaloRings
model['useShowerShape'] = useShowerShape
model['useTrack'] = useTrack
discrList.append( model )
self._logger.info('neuron = %d, sort = %d, init = %d',
info['neuron'],
info['sort'],
info['init'])
# for benchmark
# for summay in list
return discrList
RingerLayer.EM1,
RingerLayer.EM2,
RingerLayer.EM3,
RingerLayer.HAD1,
RingerLayer.HAD2,
RingerLayer.HAD3,]
from RingerCore import load,save
from RingerCore import changeExtension, ensureExtension, appendToFileName, progressbar, mkdir_p
from itertools import product
import numpy as np
if args.outputPath is None:
args.outputPath = os.path.dirname(args.inputFile)
if not os.path.isdir( args.outputPath ): mkdir_p( args.outputPath )
f = load(args.inputFile)
# Copy all metada information
baseDict = { k : f[k] for k in f.keys() if not '_etBin_' in k and not '_etaBin_' in k }
nEtBins = f['nEtBins'].item()
nEtaBins = f['nEtaBins'].item()
for etIdx, etaIdx in progressbar( product(xrange(nEtBins), xrange(nEtaBins))
, nEtBins*nEtaBins
, logger = mainLogger
, prefix = 'Juicing file '):
binDict= {k:f[k] for k in f.keys() if 'etBin_%d_etaBin_%d'%(etIdx,etaIdx) in k}
binDict.update(baseDict)
from copy import deepcopy
for layer in caloLayers:
pp=PreProcChain([RingerLayerSegmentation(layer=layer)])
tmpBinDict = deepcopy(binDict)
pprint(args.inputFiles)
## Treat special arguments
if len( args.inputFiles ) == 1:
args.inputFiles = csvStr2List( args.inputFiles[0] )
args.inputFiles = expandFolders( args.inputFiles )
mainLogger.verbose("All input files are:")
if mainLogger.isEnabledFor( LoggingLevel.VERBOSE ):
pprint(args.inputFiles)
for inFile in progressbar(args.inputFiles, len(args.inputFiles),
logger = mainLogger, prefix = "Processing files "):
# Treat output file name:
from RingerCore import checkExtension, changeExtension, load, save
if checkExtension( inFile, "tgz|tar.gz|pic" ):
cOutputName = changeExtension( inFile, '.mat' )
if args.change_output_folder:
import os.path
cOutputName = os.path.join( os.path.abspath(args.change_output_folder) , os.path.basename(cOutputName) )
data = load( inFile, useHighLevelObj = False )
from scipy.io import savemat
try:
savemat( cOutputName, data )
except ImportError:
self._logger.fatal(("Cannot save matlab file, it seems that scipy is not "
"available."), ImportError)
mainLogger.info("Successfully created matlab file: %s", cOutputName)
else:
mainLogger.error("Cannot transform files '%s' to matlab." % inFile)
# end of (for fileCollection)
def makeSummaryMonSlides(outputs, nbins, choicesfile):
from scipy.io import loadmat
f = loadmat(choicesfile)
choices = dict()
choices['Pd'] = f['choices']['Pd'][0][0]
choices['Pf'] = f['choices']['Pf'][0][0]
choices['SP'] = f['choices']['SP'][0][0]
nbinchoices = 0
for etchoices in choices['Pd']:
nbinchoices += len(etchoices)
if not nbins == nbinchoices:
raise NameError('Choices Archieve Error')
net = len(choices['Pd'])
neta = len(choices['Pd'][0])
unbinned = False
if net == 1 and neta == 1:
unbinned = True
f = load('{}/anex.pic.gz'.format(outputs + '_et0_eta0'))
benchmarkNames = f['perf'].keys()
for benchmarkName in benchmarkNames:
slideAnex = []
for et in xrange(net):
etlist = []
for eta in xrange(neta):
etaDict = dict()
neuron = choices[benchmarkName.split('_')[-1]][et][eta]
basepath = outputs
basepath += ('_et%d_eta%d') % (et, eta)
f = load('{}/anex.pic.gz'.format(basepath))
etaDict['nsignal'] = f['nsignal']
etaDict['nbackground'] = f['nbackground']
etstr = f['bounds']['etbinstr']
etastr = f['bounds']['etabinstr']
perfs = f['perf'][benchmarkName]
refBench = perfs['config_' + str(neuron).zfill(3)].getRef()
perfBench = perfs['config_' + str(neuron).zfill(3)].getPerf()
bestNetBench = perfs['config_' + str(neuron).zfill(3)].rawOp()
etaDict['sortMaximuns'] = perfs[
'config_' + str(neuron).zfill(3)].getSortPerfs()
detR = (refBench['det'])
spR = (refBench['sp'])
faR = (refBench['fa'])
detRstr = r'{:.2f}'.format(detR)
spRstr = r'{:.2f}'.format(spR)
faRstr = r'{:.2f}'.format(faR)
perf = dict()
perf['refBench'] = dict()
perf['refBench']['det'] = detR
perf['refBench']['sp'] = spR
perf['refBench']['fa'] = faR
refBench = [detRstr, spRstr, faRstr]
detP = (perfBench['detMean'], perfBench['detStd'])
spP = (perfBench['spMean'], perfBench['spStd'])
faP = (perfBench['faMean'], perfBench['faStd'])
detPstr = r'%.2f$\pm$%.2f' % (detP)
spPstr = r'%.2f$\pm$%.2f' % (spP)
faPstr = r'%.2f$\pm$%.2f' % (faP)
perf['perfBench'] = dict()
perf['perfBench']['det'] = detP
perf['perfBench']['sp'] = spP
perf['perfBench']['fa'] = faP
perfBench = [detPstr, spPstr, faPstr]
detB = (bestNetBench['det'] * 100)
spB = (bestNetBench['sp'] * 100)
faB = (bestNetBench['fa'] * 100)
detBstr = r'{:.2f}'.format(detB)
spBstr = r'{:.2f}'.format(spB)
faBstr = r'{:.2f}'.format(faB)
perf['bestNetBench'] = dict()
perf['bestNetBench']['det'] = detB
perf['bestNetBench']['sp'] = spB
perf['bestNetBench']['fa'] = faB
bestNetBench = [detBstr, spBstr, faBstr]
perfs = [refBench, perfBench, bestNetBench]
graphSections = [
'All Sorts(Validation)',
'All ROC Sorts(Validation)',
'All Sorts(Operation)',
'All ROC Sorts(Operation)',
'Best Network',
'Best Operation Output',
]
figures = [
'{}/figures/{}/neuron_{}/plot_{}_neuron_{}_sorts_val.pdf'.
format(basepath, benchmarkName, neuron, benchmarkName,
neuron),
'{}/figures/{}/neuron_{}/plot_{}_neuron_{}_sorts_roc_tst.pdf'
.format(basepath, benchmarkName, neuron, benchmarkName,
neuron),
'{}/figures/{}/neuron_{}/plot_{}_neuron_{}_sorts_op.pdf'.
format(basepath, benchmarkName, neuron, benchmarkName,
neuron),
'{}/figures/{}/neuron_{}/plot_{}_neuron_{}_sorts_roc_op.pdf'
.format(basepath, benchmarkName, neuron, benchmarkName,
neuron),
'{}/figures/{}/neuron_{}/plot_{}_neuron_{}_best_op.pdf'.
format(basepath, benchmarkName, neuron, benchmarkName,
neuron),
'{}/figures/{}/neuron_{}/plot_{}_neuron_{}_best_op_output.pdf'
.format(basepath, benchmarkName, neuron, benchmarkName,
neuron),
]
figuresDict = dict(zip(graphSections, figures))
etaDict['neuron'] = neuron
etaDict['figures'] = figuresDict
etaDict['graphSections'] = graphSections
etaDict['perfs'] = perfs
etaDict['perf'] = perf
etaDict['etastr'] = etastr
etaDict['etstr'] = etstr
etlist.append(etaDict)
# for eta
slideAnex.append(etlist)
#for et
totalPerfs = calcPerformance(slideAnex, net, neta)
with BeamerTexReportTemplate1(theme='Berlin',
_toPDF=True,
title=benchmarkName,
outputFile=benchmarkName,
font='structurebold'):
with BeamerSection(name='Performance'):
if not unbinned:
l1 = ['']
sideline = '{|c|ccc|}'
l1.extend([r'Pd(\%)', r'SP(\%)', r'PF(\%)'])
bodylines = []
la = [r'\hline' + '\n' + r'\text{CrossValidation}']
lb = ['Reference']
lc = ['bestNetBench']
refBench = [
totalPerfs['pdRef'], totalPerfs['spRef'],
totalPerfs['pfRef']
]
bestNetBench = [
totalPerfs['pdBest'], totalPerfs['spBest'],
totalPerfs['pfBest']
]
detP = (totalPerfs['pdTuning'], totalPerfs['pdStdTuning'])
spP = (totalPerfs['spTuning'], totalPerfs['spStdTuning'])
pfP = (totalPerfs['pfTuning'], totalPerfs['pfStdTuning'])
detPstr = r'%.2f$\pm$%.2f' % (detP)
spPstr = r'%.2f$\pm$%.2f' % (spP)
pfPstr = r'%.2f$\pm$%.2f' % (pfP)
perfBench = [detPstr, spPstr, pfPstr]
la.extend(perfBench)
lb.extend(refBench)
lc.extend(bestNetBench)
bodylines.extend([la, lb, lc])
linhas = []
linhas.append(l1)
linhas.extend(bodylines)
with BeamerSubSection(name='General Performance'):
BeamerTableSlide(
title='General Performance',
linhas=linhas,
sideline=sideline,
caption='Efficiences',
)
l1 = ['', '']
l2 = ['', '']
sideline = '{|c|c|'
for et in xrange(net):
sideline += 'ccc|'
l1.extend(['', slideAnex[et][0]['etstr'], ''])
l2.extend([r'Pd(\%)', r'SP(\%)', r'PF(\%)'])
sideline += '}'
etlines = []
for eta in xrange(neta):
la = [
r'\hline' + slideAnex[0][eta]['etastr'],
'CrossValidation'
]
lb = [r'', 'Reference']
lc = [r'', 'bestNetBench']
for et in xrange(net):
prfs = slideAnex[et][eta]['perfs']
refBench = prfs[0]
perfBench = prfs[1]
bestNetBench = prfs[2]
la.extend(perfBench)
lb.extend(refBench)
lc.extend(bestNetBench)
etlines.extend([la, lb, lc])
linhas = []
linhas.append(l1)
linhas.append(l2)
linhas.extend(etlines)
with BeamerSubSection(name='All Bins Performance'):
BeamerTableSlide(
title='All Bins Performance',
linhas=linhas,
sideline=sideline,
caption='Efficiences',
)
for et in xrange(net):
l1 = ['', '']
l2 = ['', '']
sideline = '{|c|c|ccc|}'
l1.extend(['', slideAnex[et][0]['etstr'], ''])
l2.extend([r'Pd(\%)', r'SP(\%)', r'PF(\%)'])
etlines = []
for eta in xrange(neta):
if unbinned:
la = [r'\hline', 'CrossValidation']
else:
la = [
r'\hline' + slideAnex[0][eta]['etastr'],
'CrossValidation'
]
lb = [r'Neuron', 'Reference']
lc = [r'', 'bestNetBench']
neuron = slideAnex[et][eta]['neuron']
prfs = slideAnex[et][eta]['perfs']
lc[0] = '%d' % neuron
refBench = prfs[0]
perfBench = prfs[1]
bestNetBench = prfs[2]
la.extend(perfBench)
lb.extend(refBench)
lc.extend(bestNetBench)
etlines.extend([la, lb, lc])
linhas = []
if not unbinned:
linhas.append(l1)
linhas.append(l2)
linhas.extend(etlines)
if unbinned:
BeamerTableSlide(
title='Performance',
linhas=linhas,
sideline=sideline,
caption='Efficiences',
)
else:
with BeamerSubSection(name=slideAnex[et][0]['etstr']):
BeamerTableSlide(
title='Performance',
linhas=linhas,
sideline=sideline,
caption='Efficiences',
)
with BeamerSection(name='Figures'):
if unbinned:
neuron = slideAnex[0][0]['neuron']
graphSections = slideAnex[0][0]['graphSections']
figures = slideAnex[0][0]['figures']
for graph in graphSections:
BeamerFigureSlide(title=graph + ' Neuron: ' +
str(neuron),
path=figures[graph])
else:
for et in xrange(net):
with BeamerSubSection(name=slideAnex[et][0]['etstr']):
for eta in xrange(neta):
neuron = slideAnex[et][eta]['neuron']
graphSections = slideAnex[et][eta][
'graphSections']
figures = slideAnex[et][eta]['figures']
with BeamerSubSubSection(
name=slideAnex[et][eta]['etastr'] +
', Neuron: {}'.format(neuron)):
for graph in graphSections:
BeamerFigureSlide(title=graph,
path=figures[graph])
return
# from TuningTools.dataframe import RingerOperation
args.operation = RingerOperation.tostring(args.operation)
nnList_calo = {}
nnList_track = {}
## Retrieving Calorimeter Networks
logger.info("Retrieving Calorimeter Networks...")
filelist = args.network_calo
if isinstance(filelist, (str)): filelist = [filelist]
elif isinstance(filelist, (tuple)): filelist = list(filelist)
for et in MatlabLoopingBounds(args.et_bins):
nnList_calo[et] = {}
for eta in MatlabLoopingBounds(args.eta_bins):
tmp = load(filelist.pop(0))
nnList_calo[et][eta] = {}
for x in ['Pd', 'Pf', 'SP']:
nnList_calo[et][eta][x] = {}
hn = tmp['OperationPoint_%s_%s' %
(opName, x)]['infoTstBest']['neuron']
logger.debug("Reference %s: %i neurons in the hidden layer" %
(x, hn))
for sort in tmp['infoPPChain'].keys():
# TODO: Add a progressbar to the loop
nnList_calo[et][eta][x][sort] = tmp[
'OperationPoint_%s_%s' %
(opName, x)]['config_%1.3i' %
(hn)][sort]['infoOpBest']['discriminator']
## Retrieving Tracking Networks
rawdict['_passed'] = mat[rowIdx][columnIdx]
def changeTotal(key, mat):
for lkey in d[key].item():
for rowIdx, array1 in enumerate(d[key].item()[lkey]):
for columnIdx, rawdict in enumerate(array1):
rawdict['_count'] = 100.
changePassed(key, mat)
changeTotal(key, mat)
from RingerCore import load, save
a = load(
'mc15_13TeV.361106.423300.sgn.trigegprobes.bkg.vetotruth.trig.l2calo.eg.std.grid.medium.npz'
)
d = dict(a)
changeOp(d, 'signal_efficiencies', mergeEffTable(veryloose20160701))
changeOp(d, 'background_efficiencies', pfrefs_vloose)
save(
d,
'mc15_13TeV.361106.423300.sgn.trigegprobes.bkg.vetotruth.trig.l2calo.eg.std.grid.veryloose',
protocol='savez_compressed')
d = dict(a)
changeOp(d, 'signal_efficiencies', mergeEffTable(loose20160701))
changeOp(d, 'background_efficiencies', pfrefs_loose)
save(
d,