R D Schaffer authored 2 years ago and Vakhtang Tsulaia committed 2 years ago

physlite updates

Rui Wang authored 2 years ago and Adam Edward Barton committed 2 years ago

add 'EventInfo.RandomRunNumber.PileupWeight_NOSYS.GenFiltHT.GenFiltMET.timeStampNSOffset' to PHYSLITE.py

Bertrand Martin Dit Latour authored 2 years ago and Walter Lampl committed 2 years ago

DerivationFrameworkPHYS: fix tau thinning in PHYSLITE and add DeepSet tau ID to PHYSLITE output

Hello,

This MR fixes a problem with tau thinning in PHYSLITE caused by !59677 : the TauTrackParticleThinning was dropped from the list of thinning tools supported in PhysCommonThinningConfig and replaced with an all-in-one tau thinning tool; this was propagated to the PHYS configuration but not to PHYSLITE.

This MR drops the TauTrackParticleThinning from PHYSLITE and only keeps the new TauThinningTool.
It fixes all the PHYSLITE ART tests that are failing today.

Besides, I have added the new Deep Set tau ID to the PHYSLITE output, that's 2 floats (scores) and 4 chars (WPs).
The new tau ID was introduced in !59725 but the output was not kept in PHYSLITE.

For completeness, there is no tau thinning issue in PHYS->PHYSLITE, and the PHYSLITE output seems to be the same (for what concerns taus) when running AOD->PHYSLITE and AOD->PHYS->PHYSLITE.

Unrelated to this MR, I also checked that the multiple tau MRs from yesterday didn't introduce differences in other objects for the PHYS format (I compared today's nightly with yesterday's nightly). I had to ignore tau branches and InDetTrackParticles for this check, as these are affected by tau thinning.

Cheers,
Bertrand

Bertrand Martin Dit Latour authored 2 years ago and Frank Winklmeier committed 2 years ago

DerivationFrameworkPhys: update EventInfo ad TauJets content

Hello,

This MR is doing the following:
- add GenFiltHT and GenFiltMET variables to EventInfo, currently they are decorated but not written out, these are needed to combine ttbar and single top inclusive samples with HT- and MET-sliced samples
- add the fixed electron-tau RNN WPs (original WPs are broken in RAWtoALL, we redecorate the fixed WPs in AOD->DAOD, these WPs are part of the tau smart slimming list but must be manually added in PHYSLITE)
- drop EventInfo.hardScatterVertexLink as this is included in the corresponding smart slimming list

Marking as urgent in case a release is to be built very soon.

Cheeers,
Bertrand

Michael Holzbock authored 2 years ago and Frank Winklmeier committed 2 years ago

adding back JVFCorr to allow recalculation of NNJvt

Zach Marshall authored 2 years ago and Frank Winklmeier committed 2 years ago

Adding NNJvtPass decoration to PHYSLITE

Following discussions in AMG, the modern recommendations will use
NNJvtPass. My understanding of the discussion thread was that the other
variables should be dropped, since they won't be needed for analyses
running on PHYSLITE anyway.

Thomas Strebler authored 2 years ago and Frank Winklmeier committed 2 years ago

FTag training run 4 in derivation

James Catmore authored 2 years ago and Tadej Novak committed 2 years ago

Draft: Combining the thinning tools for PHYS and PHYSLITE into a single config

ATLASG-2390

This MR combines the (currently separate) thinning requirements for PHYS and PHYSLITE into a common config script. This will help to keep them in synch, and simplifies the requirement that the config supports both AOD->PHYSLITE and PHYS->PHYSLITE workflows. It also reduces duplicated config code.

James Catmore authored 2 years ago and Adam Edward Barton committed 2 years ago

Adding taus to AnalysisMET calculatiom in CA version of PHYSLITE

Due to an oversight during the migration to CA, the AnalysisTaus were not passed to the AnalysisMET calculation in the CA version of PHYSLITE. This is rectified in this MR.

James Catmore authored 2 years ago and Tadej Novak committed 2 years ago

Updating trigger matching configuration in PHYS and PHYSLITE

ATR-26576

This MR makes the following adjustments to the trigger matching configuration in PHYS and PHYSLITE:

- removes run-2 style matching containers for run-3 files, instead relying on the compact navigation as explained in ATR-26576
- in PHYSLITE, adds the building of run-2 style matching containers (`AnalysisTrigMatch`) for run-2 files, which was missing previously
- adjusts the keyword arguments for some of the run-2 uonfig fragments to allow a clean use of `**kwargs` all the way down to the TriggerMatchingTool

This MR should reduce significantly the size of the trigger domain in the run 3 files.

Giovanni Marchiori authored 2 years ago and Tadej Novak committed 2 years ago

Fix bugs in EgammaTrackParticleThinning and add GSF track thinning to PHYSLITE

James Catmore authored 2 years ago and Edward Moyse committed 2 years ago

Fixing over-writing of the extra content dictionary in CA version of PHYSLITE

This fixes a bug where the "AppendToDictionary" member of the slimming helper in PHYSLITE was over-written. This led to all of the trigger matching being omitted. With this fix the trigger matching containers are added to the CA version of PHYSLITE, bringing them into line with the legacy config. It should be noted that as things stand the trigger matching containers are _far_ too big (in both config versions), so this will need to be addressed in a separate MR.

P.S. The "AppendToDictionary" should probably be made into a method (as it name kind of implies already - but falsely) to avoid this sort of thing happening, but it will require many changes across dozens of python files that are already in flux due to the CA migration, so again, best to address separately.

James Catmore authored 2 years ago and Edward Moyse committed 2 years ago

Draft: AnalysisTauJets in CA version of DAOD_PHYS

The relevant lines for including AnalysisTauJets in DAOD_PHYS are uncommented in the CA version. This works technically but it isnÂ't clear whether this configuration is safe, due to the use of public tools in the tau analysis sequence. Draft until this is clearer.

James Catmore authored 2 years ago and Adam Edward Barton committed 2 years ago

Draft: add code to allow DAOD_PHYSLITE to be made from PHYS with the component accumulator

This MR adds the small amount of extra python needed to allow PHYSLITE to be made from PHYS using the Derivation_tf transform, that is, the component accumulator. It touches three parts:

1. MuonAnalysisSequence has an extra argument allowing the internally produced MuonSelectionTool to be set to run 3 geometry should the input be data18/mc20. If this isn't done the tool crashes. This doesn't happen with AOD->PHYSLITE, presumably because the properties are inherited from the MuonCommonConfig, but this isn't used for PHYS->PHYSLITE
2. DerivationSkeleton is slightly adjusted to accommodate the different arguments, and also to create a new ConfigFlag called Input.FileType (currently can be AOD, DAOD_PHYS, EVNT)... this may be useful for other applications as well
3. PHYSLITE.py adjustment itself to put in guards to ensure that code that only works with AOD input (e.g. writing common physics content and re-setting MET association) isn't scheduled. It largely uses the new FileType ConfigFlag to achieve this. Also it sets the relevant geometry in the MuonAnalysisSequence.

With these changes, the following command works

```
Derivation_tf.py --CA --inputDAOD_PHYSFile filename.pool.root --outputD2AODFile output.pool.root --formats PHYSLITE
```

On account of the disabling of the MET association, the output is not yet usable for physics analysis. However, it still makes sense to prepare this MR to allow testing of this workflow, which will be needed during run 3.

Philip Sommer authored 2 years ago and Frank Winklmeier committed 2 years ago

adding to PHYSLITE variables for recommended electron WP

William Keaton Balunas authored 2 years ago and Frank Winklmeier committed 2 years ago

Custom-vertex MET in CA-based derivations

Jana Schaarschmidt authored 2 years ago and Johannes Elmsheuser committed 2 years ago

adding UFO jets to PHYSLITE and updating jet sequence

James Catmore authored 2 years ago and Adam Edward Barton committed 2 years ago

Adding new flavour tagging and muon variables to PHYSLITE

This MR adds the new DL1dv00 and DL1dv01 to PHYSLITE, along with the CaloMuonScore variable. All are needed by standard analysis. Changes have been applied to the legacy and new configuration files for debugging purposes.

James Catmore authored 2 years ago and Walter Lampl committed 2 years ago

Draft: migration of PHYSLITE to the component accumulator

This MR is an attempt to migrate PHYSLITE to the component accumulator. This is proving tricky because the analysis sequences aren't compatible with the component accumulator, so some wrangling is required. Concretely the following changes are required:
- The DualUseConfig can now return a new-style configurable (GaudiConfig2)
- AnaAlgSequence is given a new member, being a list to hold any GaudiConfig2 components that can't be appended directly to the sequence (which is old-style). Currently the old-style sequence isn't used directly but is instead iterated over and each component added to a new-style sequence before being added to the component accumulator
- New config fragment to allow the calo cell thininng
- PHYSLITE job options themselves