update ParticleNetAK4@HLT trainings for 2023 data-taking

[scooperstein]

ParticleNet has been retrained in CMSSW 12_6_X with the latest HLT conditions and Run-3 MC samples. This network is meant to be included in a set of trigger paths for the 2023 data-taking V2 menu. Since taus are now included by default, there is no longer a separate network included without taus. Merging this pull request will enable the deployment of the new 2023 4j2b trigger paths.

Higgs trigger contacts: @mstamenk @portalesHEP
BTV@HLT conveners: @NiclasEich

@rgerosa

[cmsbuild]

A new Pull Request was created by @scooperstein for branch master.

@smuzaffar, @aandvalenzuela, @iarspider, @clacaputo, @cmsbuild, @mandrenguyen can you please review it and eventually sign? Thanks.
@perrotta, @dpiparo, @rappoccio you are the release manager for this.
cms-bot commands are listed here

[cmsbuild]

Pull request #52 was updated.

[missirol]

assign hlt

@scooperstein , please squash this PR in a single commit to clean the git history (so as to leave the V00 files untouched).

Some context: these new files are expected to be used in the near future in the 2023 HLT pp menu (they are specified as string parameters of a few modules of the HLT pp menu; see related comments in CMSHLT-2695). This means that

• this PR does not depend on any open CMSSW PR
• it simply adds a few files that would be integrated untested (this was also done in Merge ParticleNet AK4 and AK8 models for Run3 #47 , although there some manual validation was done)
• the corresponding cmsdist update needs to be backported to 13_0_X, and until there is a new 13_0_X release with that, HLT would not be able to adopt the new cms-data files (the HLT menu expected to use these files should be integrated in CMSSW in the first week of May, so the new 13_0_X release should be in place by then) (for the integration in CMSSW, an IB with this cms-data update would be sufficient, but obviously for data-taking a new release is necessary).

[cmsbuild]

New categories assigned: hlt

@missirol,@Martin-Grunewald you have been requested to review this Pull request/Issue and eventually sign? Thanks

[missirol]

@scooperstein , please see #52 (comment) and update this PR. We can switch to the new PNet training at HLT only after this PR is is integrated in 13_0_X.

[cmsbuild]

Pull request #52 was updated.

[cmsbuild]

Pull request #52 was updated.

[cmsbuild]

Pull request #52 was updated.

[scooperstein]

Hi @missirol, all - sorry for the delay. We took a few days to update the ParticleNet@HLT AK8 model to remove a couple additional input variables that we did not use last year and were not necessary for the performance. With the training now included we use exactly the same input features as last year, such that we can use the same feature producer. I have also squashed the commits, as requested. Please let us know if there is anything further needed before merging.

[missirol]

+hlt

I tested the new cms-data files with a recent HLT menu [*], and saw no issue.

[*] Based on a customisation provided offline by @scooperstein.

#!/bin/bash

# cmsrel CMSSW_13_0_X_2023-04-18-2300
# cd CMSSW_13_0_X_2023-04-18-2300/src
# cmsenv
# git cms-addpkg RecoBTag/Combined
# git clone https://github.com/scooperstein/RecoBTag-Combined RecoBTag/Combined/data
# scram build

hltGetConfiguration /dev/CMSSW_13_0_0/GRun \
   --globaltag 126X_mcRun3_2023_forPU65_v3 \
   --mc \
   --unprescale \
   --output minimal \
   --max-events 200 \
   --input /store/mc/Run3Winter23Digi/GluGlutoHHto2B2Tau_kl-1p00_kt-1p00_c2-3p00_TuneCP5_13p6TeV_powheg-pythia8/GEN-SIM-RAW/GTv3Digi_126X_mcRun3_2023_forPU65_v3-v2/2560000/e1b2127c-442b-405e-9c19-c8429a2814e6.root \
   --eras Run3 --l1-emulator FullMC --l1 L1Menu_Collisions2023_v1_0_0_xml \
   > hlt.py

cmsRun hlt.py &> hlt1.log

cat <<@EOF >> hlt.py
process.hltParticleNetONNXJetTags.preprocess_json = 'RecoBTag/Combined/data/HLT/ParticleNetAK4/V01/preprocess.json'
process.hltParticleNetONNXJetTags.model_path = 'RecoBTag/Combined/data/HLT/ParticleNetAK4/V01/particle-net.onnx'
process.hltParticleNetONNXJetTags.flav_names = ['probtauhp', 'probtauhm', 'probb', 'probc', 'probuds', 'probg', 'ptcorr']
process.hltParticleNetDiscriminatorsJetTags.discriminators[0].denominator = ['hltParticleNetONNXJetTags:prob'+foo for foo in ['b', 'c', 'uds', 'g', 'tauhp', 'tauhm']]
process.hltParticleNetDiscriminatorsJetTags.discriminators[1].denominator = ['hltParticleNetONNXJetTags:prob'+foo for foo in ['b', 'c', 'uds', 'g', 'tauhp', 'tauhm']]
del process.hltParticleNetDiscriminatorsJetTags.discriminators[2]

process.hltParticleNetONNXJetTagsAK8.preprocess_json = 'RecoBTag/Combined/data/HLT/ParticleNetAK8/V01/preprocess.json'
process.hltParticleNetONNXJetTagsAK8.model_path = 'RecoBTag/Combined/data/HLT/ParticleNetAK8/V01/particle-net.onnx'
process.hltParticleNetONNXJetTagsAK8.flav_names = ['probHtt', 'probHtm', 'probHte', 'probHbb', 'probHcc', 'probHqq', 'probHgg', 'probQCD2hf', 'probQCD1hf', 'probQCD0hf']
process.hltParticleNetDiscriminatorsJetTagsAK8.discriminators[0].denominator = ['hltParticleNetONNXJetTagsAK8:prob'+foo for foo in ['Hbb', 'QCD2hf', 'QCD1hf', 'QCD0hf']]
process.hltParticleNetDiscriminatorsJetTagsAK8.discriminators[1].denominator = ['hltParticleNetONNXJetTagsAK8:prob'+foo for foo in ['Hcc', 'QCD2hf', 'QCD1hf', 'QCD0hf']]
process.hltParticleNetDiscriminatorsJetTagsAK8.discriminators[2].denominator = ['hltParticleNetONNXJetTagsAK8:prob'+foo for foo in ['Htt', 'QCD2hf', 'QCD1hf', 'QCD0hf']]
@EOF

cmsRun hlt.py &> hlt2.log

[missirol]

urgent

@perrotta @rappoccio , please consider merging this PR. It is needed for HLT development in 13_0_X.

The RECO signature is pending, but the added files are unrelated to the offline reconstruction.

[perrotta]

please test

• There are only new files. and I really do not expect anythyng strange from the test, but let have them tested at least once before merging

[cmsbuild]

+1

Summary: https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-91e0da/32084/summary.html
COMMIT: 9778a35
CMSSW: CMSSW_13_1_X_2023-04-21-1100/el8_amd64_gcc11
User test area: For local testing, you can use /cvmfs/cms-ci.cern.ch/week1/cms-data/RecoBTag-Combined/52/32084/install.sh to create a dev area with all the needed externals and cmssw changes.

Comparison Summary

Summary:

• You potentially added 9 lines to the logs
• Reco comparison results: 4 differences found in the comparisons
• DQMHistoTests: Total files compared: 48
• DQMHistoTests: Total histograms compared: 3460915
• DQMHistoTests: Total failures: 3
• DQMHistoTests: Total nulls: 0
• DQMHistoTests: Total successes: 3460890
• DQMHistoTests: Total skipped: 22
• DQMHistoTests: Total Missing objects: 0
• DQMHistoSizes: Histogram memory added: 0.0 KiB( 47 files compared)
• Checked 207 log files, 159 edm output root files, 48 DQM output files
• TriggerResults: no differences found

[perrotta]

+1

[perrotta]

merge