Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add Unified ParT to tau NanoAOD #44691

Merged
merged 16 commits into from
Apr 19, 2024
Merged

Conversation

lucasrussell01
Copy link
Contributor

@lucasrussell01 lucasrussell01 commented Apr 10, 2024

PR description:

This PR adds Unified ParT information to taus stored in NanoAOD.
-> These changes require the PRs from @AlexDeMoor to be merged #44641, based on the model in cms-data/RecoBTag-Combined#57

Added (for PUPPI jets):

  • UnifiedParT discriminators vs jets, electrons and muons,
  • Decay mode corresponding with the UnifiedParT tau node with the highest score,
  • Correction for tau pt,
  • Charge assignment confidence (signed),
  • UnifiedParT scores for individual decay-modes (normalised to the sum of the scores for all tau nodes).

Corresponding ParticleNet 2023 variables are stored for CHS jets.

Performance of UParT is described in: https://indico.cern.ch/event/1403350/contributions/5899772/attachments/2832957/4949984/UParT%202024%20-%20final%20results.pdf

Backport PR is here: #44723

PR validation:

Tested successfully with NanoAOD workflow using CMSSW_14_0 RELVAL file, with pending PR UnifiedParT changes and model merged.

Distributions of UParT and PNet for taus compared here:
https://indico.cern.ch/event/1401414/contributions/5891063/attachments/2838625/4961188/UParT_Integration.pdf

@cmsbuild
Copy link
Contributor

cmsbuild commented Apr 10, 2024

cms-bot internal usage

@cmsbuild
Copy link
Contributor

-code-checks

Logs: https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-44691/39893

Code check has found code style and quality issues which could be resolved by applying following patch(s)

@cmsbuild
Copy link
Contributor

-code-checks

Logs: https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-44691/39894

Code check has found code style and quality issues which could be resolved by applying following patch(s)

@cmsbuild
Copy link
Contributor

+code-checks

Logs: https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-44691/39897

@cmsbuild
Copy link
Contributor

A new Pull Request was created by @lucasrussell01 for master.

It involves the following packages:

  • PhysicsTools/NanoAOD (xpog)
  • PhysicsTools/PatAlgos (reconstruction, xpog)

@mandrenguyen, @hqucms, @jfernan2, @vlimant, @cmsbuild can you please review it and eventually sign? Thanks.
@mmarionncern, @emilbols, @AlexDeMoor, @mariadalfonso, @hatakeyamak, @demuller, @jdolen, @jdamgov, @schoef, @AnnikaStein, @Senphy, @ahinzmann, @mbluj, @Ming-Yan, @gkasieczka, @rappoccio, @azotz, @JyothsnaKomaragiri, @andrzejnovak, @nhanvtran, @gouskos, @seemasharmafnal, @gpetruc this is something you requested to watch as well.
@sextonkennedy, @antoniovilela, @rappoccio you are the release manager for this.

cms-bot commands are listed here

@hqucms
Copy link
Contributor

hqucms commented Apr 16, 2024

Thank you @lucasrussell01 !
For the performance, I was more looking for some studies performed within the TAU POG, particularly with the implementations in this PR. Could you or the @cms-sw/tau-pog-l2 point them to us?

@hqucms What in particular would you like to see?
We checked the nano outputs visually, can produce some histograms of the added variables in Nano if you like.

OK that would be good. Could you please plot both PNet and UParT so we can have also a qualitative comparison?

Hi @hqucms,

Some plots and brief comments are available here: https://indico.cern.ch/event/1401414/contributions/5891063/attachments/2838625/4961188/UParT_Integration.pdf

I will add this to the PR description.

Thanks a lot @lucasrussell01 , that's very useful to have!

Can you also update the backport PR so we can start testing it?

@lucasrussell01
Copy link
Contributor Author

Hi @mandrenguyen @hqucms @vlimant @jfernan2, just to quickly discuss these failed tests.

It seems that the RelVal errors are related to the fact that testing needs to be done with the changes in #44641 merged - as otherwise RecoBTag.ONNXRuntime.pfUnifiedParticleTransformerAK4JetTags_cfi doesn't exist. Also am I correct in thinking the the failed unit test is a general problem, and not related to this PR?

Please let me know if you need anything from our side to move forward :)

Thank you @lucasrussell01 !
For the performance, I was more looking for some studies performed within the TAU POG, particularly with the implementations in this PR. Could you or the @cms-sw/tau-pog-l2 point them to us?

@hqucms What in particular would you like to see?
We checked the nano outputs visually, can produce some histograms of the added variables in Nano if you like.

OK that would be good. Could you please plot both PNet and UParT so we can have also a qualitative comparison?

Hi @hqucms,
Some plots and brief comments are available here: https://indico.cern.ch/event/1401414/contributions/5891063/attachments/2838625/4961188/UParT_Integration.pdf
I will add this to the PR description.

Thanks a lot @lucasrussell01 , that's very useful to have!

Can you also update the backport PR so we can start testing it?

@hqucms Should be done now - thanks!

@cmsbuild
Copy link
Contributor

-1

Failed Tests: UnitTests
Summary: https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-73a430/38869/summary.html
COMMIT: a02009c
CMSSW: CMSSW_14_1_X_2024-04-16-1100/el8_amd64_gcc12
Additional Tests: NANO
User test area: For local testing, you can use /cvmfs/cms-ci.cern.ch/week1/cms-sw/cmssw/44691/38869/install.sh to create a dev area with all the needed externals and cmssw changes.

The following merge commits were also included on top of IB + this PR after doing git cms-merge-topic:

You can see more details here:
https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-73a430/38869/git-recent-commits.json
https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-73a430/38869/git-merge-result

Unit Tests

I found 1 errors in the following unit tests:

---> test runtestPhysicsToolsPatAlgos had ERRORS

Comparison Summary

Summary:

  • You potentially added 10 lines to the logs
  • ROOTFileChecks: Some differences in event products or their sizes found
  • Reco comparison results: 936 differences found in the comparisons
  • DQMHistoTests: Total files compared: 48
  • DQMHistoTests: Total histograms compared: 3319227
  • DQMHistoTests: Total failures: 231
  • DQMHistoTests: Total nulls: 0
  • DQMHistoTests: Total successes: 3318976
  • DQMHistoTests: Total skipped: 20
  • DQMHistoTests: Total Missing objects: 0
  • DQMHistoSizes: Histogram memory added: 45.55500000000001 KiB( 47 files compared)
  • DQMHistoSizes: changed ( 11634.0,... ): 3.894 KiB Physics/NanoAODDQM
  • DQMHistoSizes: changed ( 13234.0,... ): 2.033 KiB Physics/NanoAODDQM
  • Checked 202 log files, 165 edm output root files, 48 DQM output files
  • TriggerResults: no differences found

NANO Comparison Summary

Summary:

  • You potentially added 5 lines to the logs
  • ROOTFileChecks: Some differences in event products or their sizes found
  • Reco comparison results: 63 differences found in the comparisons
  • DQMHistoTests: Total files compared: 15
  • DQMHistoTests: Total histograms compared: 16456
  • DQMHistoTests: Total failures: 35
  • DQMHistoTests: Total nulls: 0
  • DQMHistoTests: Total successes: 16421
  • DQMHistoTests: Total skipped: 0
  • DQMHistoTests: Total Missing objects: 0
  • DQMHistoSizes: Histogram memory added: 41.489 KiB( 14 files compared)
  • DQMHistoSizes: changed ( 2500.001,... ): 3.894 KiB Physics/NanoAODDQM
  • DQMHistoSizes: changed ( 2500.011,... ): 2.033 KiB Physics/NanoAODDQM
  • Checked 55 log files, 32 edm output root files, 15 DQM output files

Nano size comparison Summary:

Sample kb/ev ref kb/ev diff kb/ev ev/s/thd ref ev/s/thd diff rate mem/thd ref mem/thd
2500.0 2.783 2.695 0.088 ( +3.3% ) 3.52 4.43 -20.4% 2.252 2.237
2500.001 2.897 2.805 0.092 ( +3.3% ) 3.17 4.00 -20.6% 2.776 2.650
2500.002 2.843 2.753 0.090 ( +3.3% ) 3.24 4.08 -20.5% 2.671 2.648
2500.01 1.446 1.386 0.061 ( +4.4% ) 6.06 7.99 -24.1% 2.315 2.272
2500.011 1.906 1.823 0.083 ( +4.5% ) 3.32 4.43 -25.1% 2.481 2.418
2500.012 1.761 1.687 0.074 ( +4.4% ) 4.86 6.25 -22.3% 2.467 2.351
2500.1 2.354 2.339 0.015 ( +0.6% ) 4.53 4.51 +0.5% 2.071 2.066
2500.2 2.459 2.443 0.015 ( +0.6% ) 5.11 5.20 -1.7% 1.982 1.979
2500.21 1.286 1.274 0.012 ( +0.9% ) 3.49 3.51 -0.5% 2.271 2.262
2500.211 1.668 1.653 0.015 ( +0.9% ) 3.17 3.10 +2.5% 2.351 2.365
2500.3 2.229 2.214 0.015 ( +0.7% ) 9.76 9.73 +0.2% 1.974 1.968
2500.301 2.833 2.815 0.019 ( +0.7% ) 8.27 8.38 -1.3% 1.959 1.952
2500.31 7.164 7.164 0.000 ( +0.0% ) 1.31 1.43 -8.0% 1.707 1.708
2500.311 1.568 1.568 0.000 ( +0.0% ) 6.65 6.98 -4.6% 1.056 1.058
2500.312 540.457 540.457 0.000 ( +0.0% ) 0.53 0.53 -0.4% 1.595 1.599
2500.313 817.694 817.694 0.000 ( +0.0% ) 0.72 0.74 -2.6% 1.587 1.585
2500.32 1.350 1.354 -0.004 ( -0.3% ) 12.41 12.78 -2.9% 2.367 2.367
2500.321 1.759 1.761 -0.002 ( -0.1% ) 8.46 8.92 -5.2% 2.418 2.453
2500.322 1.240 1.240 0.000 ( +0.0% ) 9.12 9.23 -1.2% 2.138 2.210
2500.323 7.772 7.772 0.000 ( +0.0% ) 3.36 3.45 -2.7% 1.928 1.921
2500.324 1.880 1.882 -0.002 ( -0.1% ) 9.06 9.30 -2.6% 2.166 2.202
2500.325 4.167 4.291 -0.124 ( -2.9% ) 4.37 4.10 +6.5% 2.169 2.202
2500.326 3.337 3.205 0.131 ( +4.1% ) 1.66 1.78 -6.5% 2.278 1.763
2500.327 1.815 1.816 -0.001 ( -0.1% ) 9.24 9.37 -1.4% 2.309 2.299
2500.4 2.374 2.388 -0.014 ( -0.6% ) 9.28 9.31 -0.4% 1.800 1.728
2500.401 1.891 1.891 0.000 ( +0.0% ) 8.31 8.33 -0.3% 1.689 1.690
2500.402 2.950 2.962 -0.012 ( -0.4% ) 8.12 8.02 +1.3% 1.802 1.877
2500.403 8.700 8.918 -0.219 ( -2.5% ) 3.02 2.80 +7.6% 1.836 1.945
2500.404 5.474 5.272 0.202 ( +3.8% ) 1.21 1.32 -7.8% 1.789 1.756
2500.405 2.860 2.872 -0.012 ( -0.4% ) 8.16 8.19 -0.3% 1.803 1.901
2500.5 5.194 5.194 0.000 ( +0.0% ) 15.87 15.76 +0.7% 1.558 1.491
2500.51 9.120 9.120 0.000 ( +0.0% ) 9.47 9.60 -1.4% 1.510 1.514

@hqucms
Copy link
Contributor

hqucms commented Apr 16, 2024

+1

@mandrenguyen
Copy link
Contributor

+reconstruction
This doesn't touch reco, only nanoAOD (despite what tests appear to show), as explained here:
#44691 (comment)

@cmsbuild
Copy link
Contributor

This pull request is fully signed and it will be integrated in one of the next master IBs (but tests are reportedly failing). This pull request will now be reviewed by the release team before it's merged. @rappoccio, @sextonkennedy, @antoniovilela (and backports should be raised in the release meeting by the corresponding L2)

@hqucms
Copy link
Contributor

hqucms commented Apr 19, 2024

type -requires-external

The cms-data PR cms-data/RecoBTag-Combined#57 is already merged.

@mandrenguyen
Copy link
Contributor

type tau

@cmsbuild cmsbuild added the tau label Apr 19, 2024
@hqucms
Copy link
Contributor

hqucms commented Apr 19, 2024

@cms-sw/orp-l2 Could you please merge this? Thanks!

@antoniovilela
Copy link
Contributor

Thanks

@antoniovilela
Copy link
Contributor

ignore tests-rejected with ib-failure

@antoniovilela
Copy link
Contributor

externals already merged

@antoniovilela
Copy link
Contributor

+1

@cmsbuild cmsbuild merged commit 513d36c into cms-sw:master Apr 19, 2024
13 of 14 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

5 participants