Implementation of ONNX model score computation and decoration for Taus (GNTau)

Add tools and wrappers to run ONNX-based models for Tau objects. Details can be found in FTag software meeting presentation, but in short: a new class and tool are added to tauRecTools along with some new variable calculator functionalities to run model inference and decorate TauJets and TauTracks, and DerivationFrameworkTau + tauRec are modified to call the tools and decorate the DAOD TauJets (if one also decorates the DeepSetID at the moment).

Not sure this is needed since the legacy configuration is going to be removed soon and by default the CA implementation is the one we should use now. (indeed don't think this is called anywhere)

Indeed it's not called anywhere, I added the non-CA stuff out of habit to be on the "complete" side.

Agreed, by the time this MR is merged, the legacy config will likely be phased out. The CA implementation is enough.

Can resolve as the file is removed in update to recent athena/main.

Not sure this is needed since the legacy configuration is going to be removed soon and by default the CA implementation is the one we should use now

As before, I added the non-CA stuff out of habit to be on the "complete" side

(same remark as above, we only need the CA implementation)

Can resolve as file is removed in update to recent athena/main.

Ok, not for now, but we should also make some tests to compare by how much the derivation file size is going to increase when adding these new quantities. If people are going to complain, are all these variables mandatory to be added or for example we can drop the Prob? (probably yes looking at the next part of the code)

Probabilities dropped after salt-pt/salt-onnx/athena-onnx consistency checks passed.

Were this "MaxTracks", "MaxClusters" and "MaxClusterDR" also features used in the training or are these just dummy/large values?

MaxTracks and MaxClusters are indeed used in training (though the limit is very rarely reached). MaxClusterDR was left-over as legacy functionality from the TauJetRNNEvaluator setup, and isn't used in the training.

added Derivation ML-Inference Reconstruction Tau changes-derivation-output main onnx labels

Tagging @ademaria, @edreyer, @prieck, @bdong and @ligang to begin with :)

Jenkins please retry a build

This merge request affects 3 packages:

• PhysicsAnalysis/DerivationFramework/DerivationFrameworkTau
• Reconstruction/tauRec
• Reconstruction/tauRecTools

Affected files list will not be printed in this case

Adding @jcatmore ,@martindl ,@goetz ,@emmat ,@adbailey ,@xiaozhon as watchers

added JetEtmiss analysis-review-required labels

CI Result FAILURE (hash a0996dcb)

	Athena	AnalysisBase	AthAnalysis
externals
cmake
make
tests

Full details available on this CI monitor view. Check the JIRA CI status board for known problems
Athena: number of compilation errors 0, warnings 1
AnalysisBase: number of compilation errors 0, warnings 1
AthAnalysis: number of compilation errors 0, warnings 1
For experts only: Jenkins output [CI-MERGE-REQUEST-EL9 4194]

Hello @ntamir, was there any progress on this?

added 1 commit

• 3da582e9 - Changes following GitLab comments

Compare with previous version

added 1 commit

• a0b6412f - Add noAux model to ConfigFlags

Compare with previous version

added 1 commit

• 2a8bb49e - Remove else{} block with evtStore call from VarCalc

Compare with previous version

General question: now that we have two models in place, can we put both in the DAOD PHYS to test them? Otherwise I would prefer to give the priority to the TauID only model without aux task

Hi @ntamir ,

Following the creation of the 24.0 branch from main, you should now decide whether this MR should target 24.0 or main, according to these guidelines agreed in the Software Weekly meeting: https://indico.cern.ch/event/1382755/attachments/2802320/4889268/BranchingGuidelines24.pdf

If you decide that this MR should target 24.0, please re-direct it by editing and changing the target branch in the drop-down menu. If it should stay in main, please indicate this as a reply to this message. Remember that all MRs going into 24.0 will also be swept into main.

Thanks!

James, John, Nick

added 2 commits

• 12e7cb70 - Resolve conflicts and rebase on current athena version
• 9a02ea25 - Merge branch 'GNTau_update' into 'master'

Compare with previous version

added 1279 commits

• 9a02ea25...93c874b8 - 1278 commits from branch atlas:main
• 9952453a - Merge branch athena:main into master

Compare with previous version

added 1 commit

• e563a3f1 - Changing WARNING to FATAL

Compare with previous version

I think if you issue a FATAL/ERROR message, it has to come together with a return StatusCode::Failure

Of course, added a "throw StatusCode**::**FAILURE;" statement following this (since this function doesn't explicitly return any SC).

Hi @ntamir! Just to let you know !68579 (merged) is now merged! I would highlight the new ConstituentsLoader interface, it might be possible for you to also reuse part of this.

From the FTAG side, we would like to be able to input clusters (as you do), so if you implemented the cluster loader as a ConstituentsLoader, FTAG would also be able to make use of this, which would be a great bonus

added 1865 commits

• e563a3f1...b31952fd - 1864 commits from branch atlas:main
• e30c1d3c - Fixing conflicts + rebase on main + tiny update

Compare with previous version

Jenkins please retry a build

This merge request affects 3 packages:

• PhysicsAnalysis/DerivationFramework/DerivationFrameworkTau
• Reconstruction/tauRec
• Reconstruction/tauRecTools

Affected files list will not be printed in this case

Adding @goetz ,@xiaozhon ,@jcatmore ,@martindl ,@emmat ,@adbailey as watchers

CI Result FAILURE (hash e30c1d3c)

	Athena	AnalysisBase	AthAnalysis
externals
cmake
make
tests

Full details available on this CI monitor view. Check the JIRA CI status board for known problems
Athena: number of compilation errors 0, warnings 0
AnalysisBase: number of compilation errors 0, warnings 0
AthAnalysis: number of compilation errors 0, warnings 0
For experts only: Jenkins output [CI-MERGE-REQUEST-EL9 7264] (remote access info)

added 3481 commits

• e30c1d3c...563802dc - 3480 commits from branch atlas:main
• 296cbdf2 - Merge branch athena:main into master

Compare with previous version

tagging also @gwatts and @ekourlit, as we would like to have this MR for the imminent PHYS production and this is adding some new variables in DAOD PHYS (not yet for PHYSLITE)

Link to the jira ticket for the PHYS production: ATLASDPD-2084

added 102 commits

• 296cbdf2...c699bb99 - 101 commits from branch atlas:main
• dc849466 - Merge branch athena:main into master

Compare with previous version

resolved all threads

added 1 commit

• be97ff64 - Final touch-ups for MR deployment, all issues resolved

Compare with previous version

added 1 commit

• f8ba7133 - Only run inference for taus with pT>13GeV, remove explicit future model reference from flags

Compare with previous version

marked this merge request as ready