Micro from700 reduced egamma #24
Conversation
|
ttbar MC before: after: |
|
JetHT data before: after: |
…loose matching for reduced conversion collection)
…nd photons with consistently relinked clusters and conversions
|
rebased and merged with the common customize functions |
|
(also, let me know to which branch I should make the pull request for 71x) |
| # | ||
| process.selectedPatJets.cut = cms.string("pt > 10") | ||
| process.selectedPatMuons.cut = cms.string("pt > 5 || isPFMuon || (pt > 3 && (isGlobalMuon || isStandAloneMuon || numberOfMatches > 0 || muonID('RPCMuLoose')))") | ||
| process.selectedPatElectrons.cut = cms.string("") | ||
| process.selectedPatTaus.cut = cms.string("pt > 20 && tauID('decayModeFinding')> 0.5") | ||
| process.selectedPatPhotons.cut = cms.string("pt > 15 && hadTowOverEm()<0.15 ") |
There was a problem hiding this comment.
we still had this selection!? looks like a bug no?
There was a problem hiding this comment.
ah no it was the final suggestion from susy
… behaviour of keeping supercluster and seed cluster for all output electrons and photons
|
Last commits almost fully restore old behaviour. Supercluster and seed cluster are now kept for all stored electrons and photons. (And detid/fraction vector in the supercluster is zeroed in case the subclusters and rechits are not being kept, as was discussed previously, but I'm not sure if it was ever implemented) Zeroing of internal information for the loose candidates is also turned back on in the slimming configs. (the change of pt>15 -> pt>14 for the photons is retained though) Re-running the event size check now. |
|
Total event size basically unchanged for ttbar (currently 30.0kB/event) Branch Name | Average Uncompressed Size (Bytes/Event) | Average Compressed Size (Bytes/Event) |
|
do you have also a 71X version of the whole thing? or the rebasing should be trivial? |
|
Rebasing should be trivial, (but there's one minor addition to be added to the ReducedEGProducer) I can do it once we're happy with the rest, just let me know which branch/release is most convenient to rebase onto. |
|
I got this In file included from /afs/ On Wed, Apr 30, 2014 at 3:32 PM, Josh Bendavid notifications@github.comwrote:
|
|
I have seen this happen ~randomly in other cases also. Just try compiling again? |
|
In fact the error is still there (here more lines)... I suspect I miss
/data/arizzi/microAOD/clean/CMSSW_7_0_0/src/RecoEgamma/EgammaPhotonProducers/src/ReducedEGProducer.cc:346:89: On Wed, Apr 30, 2014 at 5:13 PM, Josh Bendavid notifications@github.comwrote:
|
|
So I wasn't using the sparse checkout list actually, but for this error I think you just need to add RecoEgamma/EgammaTools (where I indeed changed the signature for one of the functions) |
|
still...
On Thu, May 1, 2014 at 10:33 AM, Josh Bendavid notifications@github.comwrote:
|
|
DataFormats/EgammaReco... (needed to add a setter to the SuperCluster) |
|
(and similarly in DataFormats/EgammaCandidates) |
|
Hi Josh, Ciao |
Implement a comprehensive slimming and relinking of electron and photon collections through a dedicated producer. This removes the duplication of embedded cluster information between electrons and photons.
Reduced conversion collections (with loose matching) are also produced.
Stored information is not exactly the same as before (can be tweaked further).
Current selection is
keepPhotons = cms.string("pt > 14 && hadTowOverEm()<0.15"),
relinkPhotons = cms.string("(r9()>0.8 || chargedHadronIso()<20 || chargedHadronIso()<0.3*pt())"),
keepGsfElectrons = cms.string(""),
relinkGsfElectrons = cms.string("pt>5"),
The "keep" selections control which objects are stored. The "relink" selections control for which objects the dependent clusters/rechits/conversions are to be stored in the associated reduced collections and refs relinked.
Contrary to before, the SuperCluster and seed cluster are only included if the "relink" selection is passed. (The last configuration for embedding was keeping these for all.)
There is additional information being stored through the conversions, and also the preshower clusters/rechits which were not included previously in the embedding. (Can be discussed if this is really needed).
Net effect on event size is a minor (1.5% total) reduction for ttbar and essentially no change for jetHT data. (Attached are edmEventSize dumps)
(I see that it won't merge automatically due to the last change in the configs, will rebase it tomorrow)