[12.4.X] TrackTfClassifier with Batch eval - backport of #38007

[leonardogiannini]

PR description:

Switching TrackTfClassifier from single track to batch (size 16) evaluation. The batch size is configurable and can be set to 1.

No physics change is expected (some differences at the level of the numerical precision were found when validating 38007), but the evaluation is about 3 times faster.

PR validation:

selection step speed up going from blue (old) to red (new) -> reducing a bit the overall tracking time

the physics is the same:
see e.g. http://uaf-10.t2.ucsd.edu/~legianni/batchDNNSel-PR/plots_ootb/effandfakePtEtaPhi.pdf
or http://uaf-10.t2.ucsd.edu/~legianni/batchDNNSel-PR/plots_highPurity/effandfakePtEtaPhi.pdf
for a validation comparison in TTbar with PU.

if this PR is a backport please specify the original PR and why you need to backport that PR:

The PR is a backport of #38007. It's only aimed to propagate the speed up.

[cmsbuild]

A new Pull Request was created by @leonardogiannini for CMSSW_12_4_X.

It involves the following packages:

• RecoTracker/FinalTrackSelectors (reconstruction)

@jpata, @cmsbuild, @clacaputo, @slava77 can you please review it and eventually sign? Thanks.
@felicepantaleo, @GiacomoSguazzoni, @JanFSchulte, @rovere, @VinInn, @ebrondol, @gpetruc, @mmusich, @mtosi, @dgulhan this is something you requested to watch as well.
@perrotta, @dpiparo, @qliphy you are the release manager for this.

cms-bot commands are listed here

• Backported from TrackTfClassifier with Batch evaluation of the DNN #38007

[slava77]

@cmsbuild please test

[cmsbuild]

+1

Summary: https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-fc46a5/24932/summary.html
COMMIT: 406aad7
CMSSW: CMSSW_12_4_X_2022-05-23-2300/el8_amd64_gcc10
User test area: For local testing, you can use /cvmfs/cms-ci.cern.ch/week0/cms-sw/cmssw/38053/24932/install.sh to create a dev area with all the needed externals and cmssw changes.

Comparison Summary

Summary:

• No significant changes to the logs found
• Reco comparison results: 0 differences found in the comparisons
• DQMHistoTests: Total files compared: 50
• DQMHistoTests: Total histograms compared: 3672464
• DQMHistoTests: Total failures: 224
• DQMHistoTests: Total nulls: 0
• DQMHistoTests: Total successes: 3672218
• DQMHistoTests: Total skipped: 22
• DQMHistoTests: Total Missing objects: 0
• DQMHistoSizes: Histogram memory added: 0.0 KiB( 49 files compared)
• Checked 208 log files, 45 edm output root files, 50 DQM output files
• TriggerResults: no differences found

[jpata]

+reconstruction

• an exact backport of TrackTfClassifier with Batch evaluation of the DNN #38007
• technical, speeds up reco
• DQM picks up numerical differences from the batched evaluation that we can consider to be insignificant

[cmsbuild]

This pull request is fully signed and it will be integrated in one of the next CMSSW_12_4_X IBs (tests are also fine) and once validation in the development release cycle CMSSW_12_5_X is complete. This pull request will now be reviewed by the release team before it's merged. @perrotta, @dpiparo, @qliphy (and backports should be raised in the release meeting by the corresponding L2)

[perrotta]

backport of #38007

[qliphy]

+1
tested in master