label='AK8PFPuppi',
candidates='puppi')
ca15CHSSequence = makeFatJets(process,
isData=options.isData,
label='CA15PFchs',
candidates='pfCHS')
ca15PuppiSequence = makeFatJets(process,
isData=options.isData,
label='CA15PFPuppi',
candidates='puppi')
from PandaProd.Producer.utils.setupBTag import initBTag, setupDoubleBTag
initBTag(process, '', 'packedPFCandidates', 'offlineSlimmedPrimaryVertices')
ak8CHSDoubleBTagSequence = setupDoubleBTag(process, 'packedPatJetsAK8PFchs',
'AK8PFchs', '', 'ak8')
ak8PuppiDoubleBTagSequence = setupDoubleBTag(process,
'packedPatJetsAK8PFPuppi',
'AK8PFPuppi', '', 'ak8')
ca15CHSDoubleBTagSequence = setupDoubleBTag(process, 'packedPatJetsCA15PFchs',
'CA15PFchs', '', 'ca15')
ca15PuppiDoubleBTagSequence = setupDoubleBTag(process,
'packedPatJetsCA15PFPuppi',
'CA15PFPuppi', '', 'ca15')
fatJetSequence = cms.Sequence(fatJetInitSequence + ak8CHSSequence +
ak8PuppiSequence + ca15CHSSequence +
ca15PuppiSequence + ak8CHSDoubleBTagSequence +
ak8PuppiDoubleBTagSequence +
ca15CHSDoubleBTagSequence +
ca15PuppiDoubleBTagSequence)
def makeFatJets(process, isData, label, candidates, ptMin=100.):
"""
@param label: AK8PFchs, CA15PFPuppi, etc.
"""
matches = re.match('(AK|CA)([0-9]+)PF(.+)', label)
if not matches:
raise RuntimeError('Unknown algo label ' + label)
algo = matches.group(1) + matches.group(2)
# set up radius and algoName from the input label
radius = float(matches.group(2)) * 0.1
if matches.group(1) == 'CA':
algoName = 'CambridgeAachen'
elif matches.group(1) == 'AK':
algoName = 'AntiKt'
pu = matches.group(3)
if pu == 'chs':
jecLabel = 'AK8PFchs' # regardless of jet algo
elif pu == 'Puppi':
jecLabel = 'AK8PFPuppi' # regardless of jet algo
else:
raise RuntimeError('Unknown PU mitigation ' + pu)
sdZcut, sdBeta = sdParams(radius)
sequence = cms.Sequence()
# Callable object that adds the second argument to process and sequence with label attached as suffix
addattr = AddAttr(process, sequence, label)
########################################
## REMAKE JETS ##
########################################
pfJets = addattr(
'pfJets',
ak4PFJets.clone(
doAreaFastjet=cms.bool(
False
), # This flag is turned on by default in CMSSW and makes things much slower, do not turn this on
jetAlgorithm=cms.string(algoName),
rParam=cms.double(radius),
src=cms.InputTag(candidates),
jetPtMin=cms.double(ptMin)))
pfJetsSoftDrop = addattr(
'pfJetsSoftDrop',
ak4PFJets.clone(
doAreaFastjet=cms.bool(
False
), # This flag is turned on by default in CMSSW and makes things much slower, do not turn this on
jetAlgorithm=cms.string(algoName),
rParam=cms.double(radius),
src=cms.InputTag(candidates),
jetPtMin=cms.double(ptMin),
useSoftDrop=cms.bool(True),
R0=cms.double(radius),
zcut=cms.double(sdZcut),
beta=cms.double(sdBeta),
writeCompound=cms.bool(True),
useExplicitGhosts=cms.bool(True),
jetCollInstanceName=cms.string("SubJets")))
subjets = cms.InputTag(pfJetsSoftDrop.getModuleLabel(), 'SubJets')
pfJetsPruned = addattr(
'pfJetsPruned',
ak4PFJets.clone(
doAreaFastjet=cms.bool(
False
), # This flag is turned on by default in CMSSW and makes things much slower, do not turn this on
jetAlgorithm=cms.string(algoName),
rParam=cms.double(radius),
src=cms.InputTag(candidates),
jetPtMin=cms.double(ptMin),
usePruning=cms.bool(True),
useExplicitGhosts=cms.bool(True),
writeCompound=cms.bool(True),
zcut=cms.double(0.1), # no idea if these parameters are correct
rcut_factor=cms.double(0.5),
nFilt=cms.int32(2),
jetCollInstanceName=cms.string("SubJets")))
########################################
## SUBSTRUCTURE ##
#######################################
Njettiness = addattr(
'Njettiness',
nJettinessAdder_cfi.Njettiness.clone(src=pfJets,
R0=cms.double(radius),
Njets=cms.vuint32(1, 2, 3, 4)))
sdKinematics = addattr(
'sdKinematics',
cms.EDProducer('RecoJetDeltaRValueMapProducer',
src=pfJets,
matched=pfJetsSoftDrop,
distMax=cms.double(radius),
values=cms.vstring('mass'),
valueLabels=cms.vstring('Mass')))
prunedKinematics = addattr(
'prunedKinematics',
cms.EDProducer('RecoJetDeltaRValueMapProducer',
src=pfJets,
matched=pfJetsPruned,
distMax=cms.double(radius),
values=cms.vstring('mass'),
valueLabels=cms.vstring('Mass')))
### subjet qg-tagging ###
subQGTag = addattr(
'subQGTag',
QGTagger.clone(srcJets=subjets, jetsLabel=cms.string('QGL_AK4PFchs')))
### subjet b-tagging ###
# sets up process.pfCombinedInclusiveSecondaryVertexV2BJetTags(label)Subjets (and necessary inputs)
sequence += setupBTag(process,
jetCollection=subjets,
suffix=label + 'Subjets',
vsuffix='',
tags=[
'pfCombinedInclusiveSecondaryVertexV2BJetTags',
'pfCombinedMVAV2BJetTags', 'pfDeepCSVJetTags',
'pfDeepCMVAJetTags'
])
### double-b tagging ###
sequence += setupDoubleBTag(process, isData, pfJets, jecLabel, subjets,
label, '', algo)
########################################
## MAKE PAT JETS ##
########################################
## MAIN JET ##
jecLevels = ['L1FastJet', 'L2Relative', 'L3Absolute']
if isData:
jecLevels.append('L2L3Residual')
jetCorrFactors = addattr(
'jetCorrFactors',
jetUpdater_cff.patJetCorrFactors.clone(src=pfJets,
payload=jecLabel,
levels=jecLevels,
primaryVertices=pvSource))
if not isData:
genJetMatch = addattr(
'genJetMatch',
patJetGenJetMatch.clone(src=pfJets,
maxDeltaR=radius,
matched='genJetsNoNu' + algo))
patJets = addattr(
'patJets',
_patJets.clone(jetSource=pfJets,
addJetCorrFactors=True,
addBTagInfo=True,
addAssociatedTracks=False,
addJetCharge=False,
addGenPartonMatch=False,
addGenJetMatch=(not isData),
getJetMCFlavour=False,
addJetFlavourInfo=False))
patJetsMod = addattr.last
patJetsMod.jetCorrFactorsSource = [jetCorrFactors]
patJetsMod.discriminatorSources = [
cms.InputTag('pfBoostedDoubleSVBJetTags' + label),
cms.InputTag('pfDeepDoubleBJetTags' + label, 'probQ'),
cms.InputTag('pfDeepDoubleBJetTags' + label, 'probH')
]
if not isData:
patJetsMod.genJetMatch = genJetMatch
for tau in ['tau1', 'tau2', 'tau3', 'tau4']:
patJetsMod.userData.userFloats.src.append(Njettiness.getModuleLabel() +
':' + tau)
patJetsMod.userData.userFloats.src.append(sdKinematics.getModuleLabel() +
':Mass')
patJetsMod.userData.userFloats.src.append(
prunedKinematics.getModuleLabel() + ':Mass')
patJetsMod.userData.userFloats.labelPostfixesToStrip = cms.vstring(label)
selectedPatJets = addattr(
'selectedPatJets',
jetSelector_cfi.selectedPatJets.clone(src=patJets,
cut='abs(eta) < 2.5'))
## SOFT DROP ##
patJetsSoftDrop = addattr(
'patJetsSoftDrop',
_patJets.clone(jetSource=pfJetsSoftDrop,
addJetCorrFactors=False,
addBTagInfo=False,
addAssociatedTracks=False,
addJetCharge=False,
addGenPartonMatch=False,
addGenJetMatch=False,
getJetMCFlavour=False,
addJetFlavourInfo=False))
selectedPatJetsSoftDrop = addattr(
'selectedPatJetsSoftDrop',
jetSelector_cfi.selectedPatJets.clone(src=patJetsSoftDrop,
cut='abs(eta) < 2.5'))
## GEN MATCH ##
if not isData:
genSubjetsMatch = addattr(
'genSubjetMatch',
patJetGenJetMatch.clone(src=subjets,
maxDeltaR=0.4,
matched='genJetsNoNuSoftDrop' + algo +
':SubJets'))
### SUBJETS ###
patSubjets = addattr('patSubjets',
_patJets.clone(
jetSource = subjets,
addJetCorrFactors = False,
addBTagInfo = True,
discriminatorSources = [
'pfCombinedInclusiveSecondaryVertexV2BJetTags' + label + 'Subjets',
'pfCombinedMVAV2BJetTags' + label + 'Subjets'
] + \
sum([['pfDeepCSVJetTags%sSubjets:prob%s' % (label, prob),
'pfDeepCMVAJetTags%sSubjets:prob%s' % (label, prob)]
# for prob in ['udsg', 'b', 'c', 'bb', 'cc']],
for prob in ['udsg', 'b', 'c', 'bb']],
[]),
addAssociatedTracks = False,
addJetCharge = False,
addGenPartonMatch = False,
addGenJetMatch = (not isData),
getJetMCFlavour = False,
addJetFlavourInfo = False
)
)
patSubjetsMod = addattr.last
patSubjetsMod.userData.userFloats.src.append(
cms.InputTag(subQGTag.getModuleLabel(), 'qgLikelihood'))
patSubjetsMod.userData.userFloats.labelPostfixesToStrip = cms.vstring(
label)
if not isData:
patSubjetsMod.genJetMatch = genSubjetsMatch
## MERGE SUBJETS BACK ##
jetMerger = addattr(
'jetMerger',
cms.EDProducer("BoostedJetMerger",
jetSrc=selectedPatJetsSoftDrop,
subjetSrc=patSubjets))
## PACK ##
addattr(
'packedPatJets',
cms.EDProducer("JetSubstructurePacker",
jetSrc=selectedPatJets,
distMax=cms.double(radius),
algoTags=cms.VInputTag(jetMerger),
algoLabels=cms.vstring('SoftDrop'),
fixDaughters=cms.bool(False)))
return sequence
