nTupleAnalysis

Process NANOAOD files to create skims (picoAODs) and event loop histograms.

To use ONNX Rutime to evaluate PyTorch models in CMSSW, need CMSSW 11

Login to an sl7 node:

ssh -Y @cmslpc-sl7.fnal.gov

cd ~/nobackup

cmsrel CMSSW_11_1_0_pre5

cd CMSSW_11_1_0_pre5/src

cmsenv

git cms-addpkg PhysicsTools/ONNXRuntime

git cms-merge-topic patrickbryant:MakePyBind11ParameterSetsIncludingCommandLineArguments

From a CMSSW/src release area

cmsenv

Checkout the nTupleAnalysis base class repo

git clone [email protected]:patrickbryant/nTupleAnalysis.git

git clone [email protected]:johnalison/nTupleHelperTools.git

git clone [email protected]:johnalison/TriggerEmulator.git

git clone [email protected]:patrickbryant/ZZ4b.git

For jet energy correction uncertainties we use the nanoAOD-tools package:

git clone https://github.com/cms-nanoAOD/nanoAOD-tools.git PhysicsTools/NanoAODTools

scram b ZZ4b/nTupleAnalysis

voms-proxy-init -voms cms

Run the following with --help to see what command line argument you can specify. Otherwise it runs with the default input files and default output directory.

nTupleAnalysis ZZ4b/nTupleAnalysis/scripts/nTupleAnalysis_cfg.py

Luminosity Data

When running on data a live count of the integrated luminosity that has been processed can be displayed. The luminosity per lumiblock in a given lumiMask json file is calculated with brilcalc.

First download the lumiMask:

wget https://cms-service-dqm.web.cern.ch/cms-service-dqm/CAF/certification/Collisions18/13TeV/PromptReco/Cert_314472-325175_13TeV_PromptReco_Collisions18_JSON.txt

mv Cert_314472-325175_13TeV_PromptReco_Collisions18_JSON.txt ZZ4b/lumiMasks/

Then setup the brilconda environment:

export PATH=$HOME/.local/bin:/cvmfs/cms-bril.cern.ch/brilconda/bin:$PATH

And calculate the lumi contained in the json file for a given trigger. You must specify the trigger version, this can be done by including "_v*" at the end of the trigger name. The recommended normtag can be found at https://twiki.cern.ch/twiki/bin/viewauth/CMS/TWikiLUM.

brilcalc lumi -c web --normtag /cvmfs/cms-bril.cern.ch/cms-lumi-pog/Normtags/normtag_BRIL.json -u /pb -i lumiMasks/Cert_314472-325175_13TeV_PromptReco_Collisions18_JSON.txt --hltpath "HLT_PFHT330PT30_QuadPFJet_75_60_45_40_TriplePFBTagDeepCSV_4p5_v*" --byls -o lumiMasks/brilcalc_2018_HLT_PFHT330PT30_QuadPFJet_75_60_45_40_TriplePFBTagDeepCSV_4p5.csv

The output csv file can be quite large. For 2018 it is 30MB. This is because there is one line for every lumiblock and there are O(1M) lumiblocks in a dataset. The option --byls tells brilcalc to give the lumi split by lumiblock; the default is to only specify the lumi per run.

Madgraph Studies

Generate ZZ->4b events in madgraph

cd madgraph

mg

launch mg5_ZZ4b

Convert .lhe to .root (in command line from ZZ4b directory)

gzip -d madgraph/mg5_ZZ4b/Events/run_01/unweighted_events.lhe.gz

ExRootLHEFConverter madgraph/mg5_ZZ4b/Events/run_01/unweighted_events.lhe madgraph/mg5_ZZ4b/Events/run_01/unweighted_events.root

Process events

py3 python/run.py -i madgraph/mg5_ZZ4b/Events/run_01/unweighted_events.root -o hists/mg5_ZZ4b_run_01.root

Hemisphere Analysis

To create the hemisphere library

nTupleAnalysis ZZ4b/nTupleAnalysis/scripts/nTupleAnalysis_cfg.py -i ZZ4b/fileLists/data18.txt -o $PWD -y 2018 --histogramming 1 --histFile hists.root --nevents -1 --createHemisphereLibrary

To analyze the hemisphere data (may take significant amount of time)

time hemisphereAnalysis ZZ4b/nTupleAnalysis/scripts/hemisphereAnalysis_cfg.py -i '$PWD/data18/hemiSphereLib_3TagEvents_*root' -o $PWD --histFile hists_3tag.root -n -1

To load the library and create mixed eventsx

nTupleAnalysis ZZ4b/nTupleAnalysis/scripts/nTupleAnalysis_cfg.py -i ZZ4b/fileLists/data18.txt -o $PWD -y 2018 --histogramming 1 --histFile hists.root --nevents -1 --loadHemisphereLibrary --inputHLib3Tag '$PWD/data18/hemiSphereLib_3TagEvents_*root' --inputHLib4Tag '$PWD/data18/hemiSphereLib_4TagEvents_*root'

Grid Pack Production

ZZ

https://twiki.cern.ch/twiki/bin/view/CMS/QuickGuideMadGraph5aMCatNLO

cd /uscms_data/d3/bryantp/

mkdir MG5_gridpack

cd MG5_gridpack

git clone https://github.com/cms-sw/genproductions.git

cd genproductions/bin/MadGraph5_aMCatNLO

./gridpack_generation.sh ZZTo4B01j_5f_NLO_FXFX cards/production/2017/13TeV/ZZTo4B01j_5f_NLO_FXFX/

ZH

must be done on lxplus because of how the condor script works (uses shared afs file system) https://twiki.cern.ch/twiki/bin/viewauth/CMS/PowhegBOXPrecompiled

cd /uscms_data/d3/bryantp/

mkdir powheg_gridpack

cd powheg_gridpack

check scram arch

echo $SCRAM_ARCH

slc7_amd64_gcc820 # twiki says to use CMSSW_11_0_1

cmsrel CMSSW_11_0_1

cd CMSSW_11_0_1/src

cmsenv

git clone https://github.com/cms-sw/genproductions.git

cd genproductions/bin/Powheg

voms

mkdir ZH

cp production/2017/13TeV/Higgs/HZJ_HanythingJ_NNPDF31_13TeV/HZJ_HanythingJ_NNPDF31_13TeV_M125_Vhadronic.input ZH/

./run_pwg_condor.py -p f -i ZH/HZJ_HanythingJ_NNPDF31_13TeV_M125_Vhadronic.input -m HZJ -f ZH -q workday -n 1000

mkdir ggZH

cp production/2017/13TeV/Higgs/ggHZ_HanythingJ_NNPDF31_13TeV/ggHZ_HanythingJ_NNPDF31_13TeV_M125_Vhadronic.input ggZH/

./run_pwg_condor.py -p f -i ggZH/ggHZ_HanythingJ_NNPDF31_13TeV_M125_Vhadronic.input -m ggHZ -f ggZH -q workday -n 1000