Merge pull request #46191 from nurfikri89/from142X202409302300_subjetbtag_puId

[MiniAOD, NanoAOD] Update AK8 subjet b-taggers, add pileup jet ID

Author: cmsbuild

PhysicsTools/NanoAOD/python/jetsAK4_Puppi_cff.py

@@ -90,6 +90,7 @@
         UParTAK4RegPtRawCorr = Var("?bDiscriminator('pfUnifiedParticleTransformerAK4JetTags:ptcorr')>0?bDiscriminator('pfUnifiedParticleTransformerAK4JetTags:ptcorr'):-1",float,precision=10,doc="UnifiedParT universal flavor-aware visible pT regression (no neutrinos), correction relative to raw jet pT"),
         UParTAK4RegPtRawCorrNeutrino = Var("?bDiscriminator('pfUnifiedParticleTransformerAK4JetTags:ptnu')>0?bDiscriminator('pfUnifiedParticleTransformerAK4JetTags:ptnu'):-1",float,precision=10,doc="UnifiedParT universal flavor-aware pT regression neutrino correction, relative to visible. To apply full regression, multiply raw jet pT by both UParTAK4RegPtRawCorr and UParTAK4RegPtRawCorrNeutrino."),
         UParTAK4RegPtRawRes = Var("?(bDiscriminator('pfUnifiedParticleTransformerAK4JetTags:ptreshigh')+bDiscriminator('pfUnifiedParticleTransformerAK4JetTags:ptreslow'))>0?0.5*(bDiscriminator('pfUnifiedParticleTransformerAK4JetTags:ptreshigh')-bDiscriminator('pfUnifiedParticleTransformerAK4JetTags:ptreslow')):-1",float,precision=10,doc="UnifiedParT universal flavor-aware jet pT resolution estimator, (q84 - q16)/2"),
+        puIdDisc = Var("userFloat('pileupJetIdPuppi:fullDiscriminant')", float,doc="Pileup ID BDT discriminant with 133X Winter24 PuppiV18 training",precision=10),
         hfsigmaEtaEta = Var("userFloat('hfsigmaEtaEta')",float,doc="sigmaEtaEta for HF jets (noise discriminating variable)",precision=10),
         hfsigmaPhiPhi = Var("userFloat('hfsigmaPhiPhi')",float,doc="sigmaPhiPhi for HF jets (noise discriminating variable)",precision=10),
         hfcentralEtaStripSize = Var("userInt('hfcentralEtaStripSize')", int, doc="eta size of the central tower strip in HF (noise discriminating variable)"),
@@ -134,6 +135,12 @@
     hfEmEF = None
 )
 
+run3_nanoAOD_pre142X.toModify(
+    jetPuppiTable.variables,
+    puIdDisc = None,
+)
+
+
 #jets are not as precise as muons
 jetPuppiTable.variables.pt.precision=10
 

PhysicsTools/NanoAOD/python/jetsAK8_cff.py

@@ -246,7 +246,8 @@ def nanoAOD_addDeepInfoAK8(process, addDeepBTag, addDeepBoostedJet, addDeepDoubl
     name = cms.string("SubJet"),
     doc  = cms.string("slimmedJetsAK8PFPuppiSoftDropPacked::SubJets, i.e. soft-drop subjets for ak8 fat jets for boosted analysis"),
     variables = cms.PSet(P4Vars,
-        btagDeepB = Var("bDiscriminator('pfDeepCSVJetTags:probb')+bDiscriminator('pfDeepCSVJetTags:probbb')",float,doc="DeepCSV b+bb tag discriminator",precision=10),
+        btagDeepFlavB = Var("bDiscriminator('pfDeepFlavourJetTags:probb')+bDiscriminator('pfDeepFlavourJetTags:probbb')+bDiscriminator('pfDeepFlavourJetTags:problepb')",float,doc="DeepJet b+bb+lepb tag discriminator",precision=10),
+        btagUParTAK4B = Var("?bDiscriminator('pfUnifiedParticleTransformerAK4DiscriminatorsJetTags:BvsAll')>0?bDiscriminator('pfUnifiedParticleTransformerAK4DiscriminatorsJetTags:BvsAll'):-1",float,precision=10,doc="UnifiedParT b vs. udscg"),
         rawFactor = Var("1.-jecFactor('Uncorrected')",float,doc="1 - Factor to get back to raw pT",precision=6),
         area = Var("jetArea()", float, doc="jet catchment area, for JECs",precision=10),
         tau1 = Var("userFloat('NjettinessAK8Subjets:tau1')",float, doc="Nsubjettiness (1 axis)",precision=10),
@@ -268,6 +269,14 @@ def nanoAOD_addDeepInfoAK8(process, addDeepBTag, addDeepBoostedJet, addDeepDoubl
     area = None,
 )
 
+run3_nanoAOD_pre142X.toModify(
+    subJetTable.variables,
+    btagDeepFlavB = None,
+    btagUParTAK4B = None,
+    btagDeepB = Var("bDiscriminator('pfDeepCSVJetTags:probb')+bDiscriminator('pfDeepCSVJetTags:probbb')",float,doc="DeepCSV b+bb tag discriminator",precision=10),
+)
+
+
 #jets are not as precise as muons
 fatJetTable.variables.pt.precision=10
 subJetTable.variables.pt.precision=10

PhysicsTools/NanoAOD/python/nanoDQM_cff.py

@@ -222,6 +222,10 @@ def _match(name):
     and 'hfHEF' not in plot.name.value() and 'hfEmEF' not in plot.name.value():
         _Jet_EarlyRun3_plots.append(plot)
 
+_Jet_pre142X_plots = cms.VPSet()
+for plot in nanoDQM.vplots.Jet.plots:
+    if 'puIdDisc' not in plot.name.value():
+        _Jet_pre142X_plots.append(plot)
 
 _SubJet_Run2_plots = cms.VPSet()
 for plot in nanoDQM.vplots.SubJet.plots:
@@ -234,6 +238,14 @@ def _match(name):
     if 'area' not in plot.name.value():
         _SubJet_EarlyRun3_plots.append(plot)
 
+_SubJet_pre142X_plots = cms.VPSet()
+for plot in nanoDQM.vplots.SubJet.plots:
+    if 'btagDeepFlavB' not in plot.name.value() and 'btagUParTAK4B' not in plot.name.value():
+        _SubJet_pre142X_plots.append(plot)
+_SubJet_pre142X_plots.extend([
+    Plot1D('btagDeepB', 'btagDeepB', 20, -1, 1, 'Deep B+BB btag discriminator'),
+])
+
 run2_nanoAOD_ANY.toModify(
     nanoDQM.vplots.FatJet,
     plots = _FatJet_Run2_plots
@@ -256,6 +268,13 @@ def _match(name):
     plots = _SubJet_EarlyRun3_plots
 )
 
+run3_nanoAOD_pre142X.toModify(
+    nanoDQM.vplots.Jet,
+    plots = _Jet_pre142X_plots
+).toModify(
+    nanoDQM.vplots.SubJet,
+    plots = _SubJet_pre142X_plots
+)
 
 _Pileup_pre13X_plots = cms.VPSet()
 for plot in nanoDQM.vplots.Pileup.plots:

PhysicsTools/NanoAOD/python/nanoDQM_cfi.py

@@ -448,6 +448,7 @@
                 Plot1D('PNetRegPtRawCorr', 'PNetRegPtRawCorr', 20, 0, 2, 'ParticleNet visible pT regression, correction relative to raw pT'),
                 Plot1D('PNetRegPtRawCorrNeutrino', 'PNetRegPtRawCorrNeutrino', 20, 0, 2, 'ParticleNet neutrino pT correction, relative to regressed visible pT'),
                 Plot1D('PNetRegPtRawRes', 'PNetRegPtRawRes', 20, 0, 0.5, 'ParticleNet per-jet resolution estimator: (q84 - q16)/2'),
+                Plot1D('puIdDisc', 'puIdDisc', 20, -1., 1., 'Pileup ID BDT discriminant with 133X Winter24 PuppiV18 training'),
                 Plot1D('chEmEF', 'chEmEF', 20, 0, 1, 'charged Electromagnetic Energy Fraction'),
                 Plot1D('chFPV0EF', 'chFPV0EF', 20, 0, 2, 'charged fromPV==0 Energy Fraction (energy excluded from CHS jets). Previously called betastar.'),
                 Plot1D('chHEF', 'chHEF', 20, 0, 2, 'charged Hadron Energy Fraction'),
@@ -815,8 +816,8 @@
             sels = cms.PSet(),
             plots = cms.VPSet(
                 Count1D('_size', 9, -0.5, 8.5, 'slimmedJetsAK8PFPuppiSoftDropPacked::SubJets, i.e. soft-drop subjets for ak8 fat jets for boosted'),
-                Plot1D('btagCMVA', 'btagCMVA', 20, -1, 1, 'CMVA V2 btag discriminator'),
-                Plot1D('btagDeepB', 'btagDeepB', 20, -1, 1, 'Deep B+BB btag discriminator'),
+                Plot1D('btagDeepFlavB', 'btagDeepFlavB', 20, -1, 1, 'DeepJet b+bb+lepb tag discriminator'),
+                Plot1D('btagUParTAK4B', 'btagUParTAK4B', 20, -1, 1, 'UnifiedParT b vs. udscg'),
                 Plot1D('eta', 'eta', 20, -4, 4, 'eta'),
                 Plot1D('hadronFlavour', 'hadronFlavour', 6, -0.5, 5.5, 'flavour from hadron ghost clustering'),
                 Plot1D('mass', 'mass', 20, -200, 200, 'mass'),

PhysicsTools/PatAlgos/python/slimming/applySubstructure_cff.py

@@ -63,15 +63,26 @@ def applySubstructure( process, postfix="" ) :
         jetCorrections = ('AK8PFPuppi', ['L2Relative', 'L3Absolute'], 'None'),
         getJetMCFlavour = False # jet flavor disabled
     )
+
     ## PATify soft drop subjets
+    from RecoBTag.ONNXRuntime.pfUnifiedParticleTransformerAK4_cff import _pfUnifiedParticleTransformerAK4JetTagsAll as pfUnifiedParticleTransformerAK4JetTagsAll
+    _btagDiscriminatorsSubjets = cms.PSet(
+      names=cms.vstring(
+        'pfDeepFlavourJetTags:probb',
+        'pfDeepFlavourJetTags:probbb',
+        'pfDeepFlavourJetTags:problepb',
+        'pfUnifiedParticleTransformerAK4DiscriminatorsJetTags:BvsAll'
+      )
+    )
+
     addJetCollection(
         process,
         postfix=postfix,
         labelName = 'AK8PFPuppiSoftDropSubjets',
         jetSource = cms.InputTag('ak8PFJetsPuppiSoftDrop'+postfix,'SubJets'),
         algo = 'ak',  # needed for subjet flavor clustering
         rParam = 0.8, # needed for subjet flavor clustering
-        btagDiscriminators = ['pfDeepCSVJetTags:probb', 'pfDeepCSVJetTags:probbb', 'pfCombinedInclusiveSecondaryVertexV2BJetTags','pfCombinedMVAV2BJetTags'],
+        btagDiscriminators = _btagDiscriminatorsSubjets.names.value(),
         jetCorrections = ('AK4PFPuppi', ['L2Relative', 'L3Absolute'], 'None'),
         explicitJTA = True,  # needed for subjet b tagging
         svClustering = True, # needed for subjet b tagging
@@ -80,14 +91,6 @@ def applySubstructure( process, postfix="" ) :
         groomedFatJets=cms.InputTag('ak8PFJetsPuppiSoftDrop') # needed for subjet flavor clustering
     )
 
-    from Configuration.Eras.Modifier_run3_common_cff import run3_common
-    run3_common.toModify(process.patJetsAK8PFPuppiSoftDropSubjets,
-                         discriminatorSources = cms.VInputTag(
-                            cms.InputTag("pfDeepCSVJetTagsAK8PFPuppiSoftDropSubjets","probb"),
-                            cms.InputTag("pfDeepCSVJetTagsAK8PFPuppiSoftDropSubjets","probbb")
-                         )
-    )
-
     # add groomed ECFs and N-subjettiness to soft dropped pat::Jets for fat jets and subjets
     process.load('RecoJets.JetProducers.ECF_cff')
     addToProcessAndTask('nb1AK8PuppiSoftDrop'+postfix, process.ecfNbeta1.clone(src = cms.InputTag("ak8PFJetsPuppiSoftDrop"+postfix), cuts = cms.vstring('', '', 'pt > 250')), process, task)