import numpy as np
import operator
# make predictions to be used with keras.predict
def predict_inputs( event, sample, jet_lstm = False):
flat_variables = np.array([[getattr( event, mva_variable) for mva_variable, _ in mva_variables]])
if jet_lstm:
lstm_jets_maxN = 10 #remove after retraining
jet_vector_var = mva_vector_variables["mva_Jet"]
jets = mva_vector_variables["mva_Jet"]["func"](event,sample=None)
jets = [ [ operator.itemgetter(varname)(jet) for varname in lstm_jetVarNames] for jet in jets[:lstm_jets_maxN] ]
# zero padding
jets += [ [0.]*len(lstm_jetVarNames)]*(max(0, lstm_jets_maxN-len(jets)))
jets = np.array([jets])
return [ flat_variables, jets ]
else:
return flat_variables
#define training samples for multiclassification
from tWZ.samples.nanoTuples_RunII_nanoAODv6_private_postProcessed import *
#use only Summer16
training_samples = [ Summer16.TWZ_NLO_DR, Summer16.TTZ ]
assert len(training_samples)==len(set([s.name for s in training_samples])), "training_samples names are not unique!"
# training selection
from tWZ.Tools.cutInterpreter import cutInterpreter
selectionString = cutInterpreter.cutString( 'trilepT-onZ1-btag1-njet3p' )
# selectionString = cutInterpreter.cutString( 'trilepT-minDLmass12-onZ1-njet4p-btag1' )
args = argParser.parse_args()
#Logger
import tWZ.Tools.logger as logger
logger = logger.get_logger("INFO", logFile=None)
import Analysis.Tools.logger as logger_an
logger_an = logger_an.get_logger("INFO", logFile=None)
if args.plot_directory == None:
args.plot_directory = plot_directory
if args.selection == None:
selectionString = "(1)"
else:
selectionString = cutInterpreter.cutString(args.selection)
# Samples
#from tWZ.samples.nanoTuples_RunII_nanoAODv6_private_postProcessed import *
from tWZ.samples.nanoTuples_Summer16_nanoAODv6_private_postProcessed import *
signal = WZ
#signal.reduceFiles(factor=20)
#TTZ.reduceFiles(factor=3)
# TTZ
backgrounds = [TTZ]
samples = backgrounds + [signal]
for sample in samples:
sample.setSelectionString(selectionString)
#elif args.era == "Run2018":
# mc = [Autumn18.TWZ_NLO_DR, Autumn18.TTZ, Autumn18.TTX_rare, Autumn18.TZQ, Autumn18.WZ, Autumn18.triBoson, Autumn18.ZZ, Autumn18.nonprompt_3l]
#elif args.era == "RunII":
# mc = [TWZ_NLO_DR, TTZ, TTX_rare, TZQ, WZ, triBoson, ZZ, nonprompt_3l]
bins = [
(2.5,5, -1.2, 1.2),
(5.,10., -1.2, 1.2),
(10.,999., -1.2, 1.2),
]
triggerSelection = '('+"||".join(triggers)+')'
leptonSelection = 'n%s_looseHybridIso==1'%args.mode
jetSelection = 'Sum$(Jet_pt>40&&abs(Jet_eta)<2.4&&JetGood_cleaned_%s_looseHybridIso)>=1'%args.mode
if args.selection:
selection = cutInterpreter.cutString(args.selection).replace("mT", "%s_looseHybridIso_mT"%args.mode)
else:
selection = "(1)"
max_events = -1
if args.small:
for sample in [data_sample] + mc:
sample.normalization = 1.
#sample.reduceFiles( factor = 10 )
sample.reduceFiles( to=3 )
#sample.scale /= sample.normalization
max_events = 30000
# Text on the plots
tex = ROOT.TLatex()
tex.SetNDC()
mc[1].style = styles.fillStyle(mc[1].color)
else:
for sample in mc: sample.style = styles.fillStyle(sample.color)
for sample in mc:
sample.read_variables = read_variables_MC
sample.setSelectionString([getLeptonSelection(mode)])
sample.weight = lambda event, sample: event.reweightBTag_SF*event.reweightPU*event.reweightL1Prefire*event.reweightTrigger#*event.reweightLeptonSF
if args.shape:
stack = Stack([mc[1]], [mc[0]])
else:
stack = Stack(mc)
# Use some defaults
Plot.setDefaults(stack = stack, weight = staticmethod(weight_), selectionString = cutInterpreter.cutString(args.selection))
################################################################################
# Now define the plots
plots = []
plots.append(Plot(
name = "N_GenJets_b",
texX = 'Number of b GenJets', texY = 'Number of Events',
attribute = lambda event, sample: event.nbjet,
binning=[11,-0.5,10.5],
))
plots.append(Plot(
name = "N_GenJets_light",
for sample in mc:
sample.read_variables = read_variables_MC
sample.weight = lambda event, sample: event.reweightBTag_SF * event.reweightPU * event.reweightL1Prefire * event.reweightTrigger #*event.reweightLeptonSF
#yt_TWZ_filter.scale = lumi_scale * 1.07314
if not args.noData:
stack = Stack(mc, data_sample)
else:
stack = Stack(mc)
# Use some defaults
Plot.setDefaults(stack=stack,
weight=staticmethod(weight_),
selectionString="(" +
cutInterpreter.cutString(args.selection) + ")&&(" +
getLeptonSelection(mode) + ")")
plots = []
plots.append(
Plot(
name='yield',
texX='',
texY='Number of Events',
attribute=lambda event, sample: 0.5 + i_mode,
binning=[4, 0, 4],
))
for name, has_lstm, model in models:
#print has_lstm, flat_variables, lstm_jets
lstm_jets_maxN = 10 #remove after retraining
jet_vector_var = mva_vector_variables["mva_Jet"]
jets = mva_vector_variables["mva_Jet"]["func"](event, sample=None)
jets = [[
operator.itemgetter(varname)(jet) for varname in lstm_jetVarNames
] for jet in jets[:lstm_jets_maxN]]
# zero padding
jets += [[0.] * len(lstm_jetVarNames)] * (max(
0, lstm_jets_maxN - len(jets)))
jets = np.array([jets])
return [flat_variables, jets]
else:
return flat_variables
#define training samples for multiclassification
from tWZ.samples.nanoTuples_RunII_nanoAODv6_private_postProcessed import *
#use only Summer16
training_samples = [Summer16.TWZ_NLO_DR, Summer16.TTZ]
assert len(training_samples) == len(set(
[s.name
for s in training_samples])), "training_samples names are not unique!"
# training selection
from tWZ.Tools.cutInterpreter import cutInterpreter
# selectionString = cutInterpreter.cutString( 'trilepT-onZ1-btag1-njet3p' )
selectionString = cutInterpreter.cutString(
'trilepT-minDLmass12-onZ1-njet4p-deepjet1')
]
weightstring = "*".join(nominalMCWeights)
lumistring = "(" + "+".join(
["%s*(year==%s)" % (l / 1000, y)
for (y, l) in lumi_year.iteritems()]) + ")"
weightstring += "*" + lumistring
weightstring_data = "weight"
Nevents = {}
for sample in mc:
yieldlist = [
sample.getYieldFromDraw(selectionString=None,
weightString=weightstring)['val'],
sample.getYieldFromDraw(
selectionString=cutInterpreter.cutString("trilepT"),
weightString=weightstring)['val'],
sample.getYieldFromDraw(selectionString=cutInterpreter.cutString(
"trilepT-minDLmass12"),
weightString=weightstring)['val'],
sample.getYieldFromDraw(selectionString=cutInterpreter.cutString(
"trilepT-minDLmass12-onZ1"),
weightString=weightstring)['val'],
sample.getYieldFromDraw(selectionString=cutInterpreter.cutString(
"trilepT-minDLmass12-onZ1-njet4p"),
weightString=weightstring)['val'],
sample.getYieldFromDraw(selectionString=cutInterpreter.cutString(
"trilepT-minDLmass12-onZ1-njet4p-btag1"),
weightString=weightstring)['val']
]
Nevents[sample.name] = yieldlist
# loop over modes
for mode in modes:
yields[mode] = {}
data_sample.texName = "data"
data_sample.setSelectionString([getLeptonSelection(mode)])
data_sample.name = "data"
data_sample.style = styles.errorStyle(ROOT.kBlack)
lumi_scale = data_sample.lumi/1000
for sample in mc: sample.style = styles.fillStyle(sample.color)
for sample in mc:
sample.setSelectionString([getLeptonSelection(mode)])
# Use some defaults
Plot.setDefaults( selectionString = cutInterpreter.cutString(args.selection) )
# if we're running a variation specify
if args.variation is not None:
selectionModifier = variations[args.variation]['selectionModifier']
mc_weight = MC_WEIGHT( variation = variations[args.variation], returntype='func')
else:
selectionModifier = None
mc_weight = None
# Stack
stack_mc = Stack( mc )
stack_data = Stack( data_sample )
############################################################################
