Tau reconstruction based on miniAOD event content #22594
Conversation
…, done (PFTau re-reconstruction the same)
…jet instead of PFJet
…te to Candidate and PFJet to BaseJet in PFTau
…taus are made from packed candidates
…sed on work by Michal Bluj), and adapt MiniAOD tau sequence accordingly
…sed on work by Michal Bluj), and adapt MiniAOD tau sequence accordingly
…sed on work by Michal Bluj), and adapt MiniAOD tau sequence accordingly
Tau reconstruction on MiniAOD input
|
@Martin-Grunewald @alja @santocch @andrius-k could you please check this PR? |
|
+1 |
|
+1 |
|
+1 |
|
@santocch the analysis-only part is restricted to two files, could you please check this PR? I would like to finally move it forward asap |
|
+1 this PR is plenty of merges, which is not the ideal way of updating, but from integration tests it dos not seem to involve any previously merged PR, and the history looks preserved in a reasonable way |
|
merge @santocch this is blocking other fixes, I move forward with the integration, the PhysicsTools look reasonable to me. Please check it in detail and sign it or propose further updates to be followed in a separate PR |
|
We would like to understand whether this functionality could - technically - be backported to 102X, or whether it requires other changes that are not present there. |
| const reco::PFCandidatePtr& pfCand = isolationPFCands.at(iPFCand); | ||
| std::cout << " pfCand #" << iPFCand << " (" << pfCand.id() << ":" << pfCand.key() << "):" | ||
| << " Pt = " << pfCand->pt() << ", eta = " << pfCand->eta() << ", phi = " << pfCand->phi() << std::endl; | ||
| const auto& isolationCands = tauRef_->isolationCands(); |
There was a problem hiding this comment.
@perrotta @slava77 @steggema after merging this PR there is a list of massive failures in the clang build in this class:
Begin processing the 14th record. Run 1, Event 14, LumiSection 1 on stream 2 at 07-Apr-2019 10:29:06.605 CEST
A fatal system signal has occurred: segmentation violation
The following is the call stack containing the origin of the signal.
Sun Apr 7 10:29:06 CEST 2019
Current Modules:
Module: RecoTauCleaner:hpsPFTauProducerSansRefs (crashed)
Module: PoolOutputModule:AODSIMoutput
Module: RPCDigiProducer:simMuonRPCDigis
Module: FastSimProducer:fastSimProducer
A fatal system signal has occurred: segmentation violation
|
+1 |
cmsbuild
added
analysis-approved
fully-signed
and removed
analysis-pending
pending-signatures
labels
|
This pull request is fully signed and it will be integrated in one of the next master IBs (tests are also fine). This pull request will be automatically merged. |
|
I was able to reproduce the crash under clang and got a traceback |
|
The fix in in #26381 |
Hi all,
this is a pull request concerned with tau reconstruction based on miniAOD event content. There has been a related PR (*) in December. After some restructuring, following proposals by Slava, this is the next iteration to integrate this feature. The restructuring required changes of data formats to be a accepted by plugins and other data formats used in TauReco, which imply small/minimal changes in quite a large number of files. The original work was initiated by @mbluj and continued and brought to completion by @steggema. An independent deep validation has been performed by @mbluj in addition to the usual checks that we apply. All tests went out fine.
We have checked and expect that the default workflows based on AOD input remain unchanged. Small differences are observed and expected when comparing taus reconstructed based on AOD input to taus reconstructed based on miniAOD input. Also for the availability of the full performance of the tau-ID (e.g. concerning discrimination against muons) a few pieces of information are still missing on the miniAOD event content. This work is left to follow-up implementations.
A presentation by @steggema is anticipated in then next Reco/AT meeting. For more details before this presentation will be given, please refer to @steggema and @mbluj.
Cheers,
Roger
(*)
#21757