AFT-646 FTAG: Decorator regex fix

• Fixing the bug introduced in the MR!60007
• Reverted the commit 65695701 that reverted the MR!60007 and then added the fix
• git revert -m 1 656957011f41e78145eed3b80766d3b44502051e (this is the command I used to revert, hope that's the right one)

---

Related JIRA issue AFT-646

---

Prior to the MR!60007, we were creating the decorators for the GN models using lwtnn::config, reading from the ONNX model metadata. The MR!60007 introduced the option to create the decorators from the ONNX model directly by introducing the GNNConfig

GNNConfig was missing the addition of Flip and Neg suffixes to the decorator names when needed. (In the old code, we were doing it while creating the getters; forgot to copy it to the code that creates the new decorators)

---

tagging @bdong @ligang @dguest

mentioned in merge request !64034 (closed)

@dguest, I realized the easiest fix, for now, is to have the regex code for GN1/2 in the createGNDecorator function. So that's what I ended up going for. I guess we can move the regex stuff outside later (when we move the getters to the GNNConfig as well)

Let me know if it looks good. I'll mark it ready.

Thanks! @nkakati , it looks good to me.

added BTagging label

changed title from Draft: FTAG: Decorator regex fix to Draft: {+https://its.cern.ch/jira/browse/AFT-646 +}FTAG: Decorator regex fix

changed title from Draft: {-https://its.cern.ch/jira/browse/-}AFT-646 FTAG: Decorator regex fix to Draft: AFT-646 FTAG: Decorator regex fix

Just adding a note here, not really relevant to merging this in: This whole fiasco could have been avoided if I'd properly locked the decorators when we added them to the container. I only recently started trying to work out how to do this and ran into some issues (see ATEAM-909), but I should get back on it. It's way too easy for us to schedule multiple algorithms that overwrite each other, and the behavior when we run the avalanche scheduler isn't well defined so far as I know.

added 1 commit

• 45fbe849 - added getFlipConverters

Compare with previous version

This looks good. Did you try producing a 10 event DAOD with / without this change and checking with acmd diff-root? If we've checked that and there are no changes we should be fine.

marked this merge request as ready

This merge request affects 1 package:

• PhysicsAnalysis/JetTagging/FlavorTagDiscriminants

This merge request affects 7 files:

• PhysicsAnalysis/JetTagging/FlavorTagDiscriminants/FlavorTagDiscriminants/DataPrepUtilities.h
• PhysicsAnalysis/JetTagging/FlavorTagDiscriminants/FlavorTagDiscriminants/GNN.h
• PhysicsAnalysis/JetTagging/FlavorTagDiscriminants/FlavorTagDiscriminants/GNNConfig.h
• PhysicsAnalysis/JetTagging/FlavorTagDiscriminants/FlavorTagDiscriminants/OnnxUtil.h
• PhysicsAnalysis/JetTagging/FlavorTagDiscriminants/Root/DataPrepUtilities.cxx
• PhysicsAnalysis/JetTagging/FlavorTagDiscriminants/Root/GNN.cxx
• PhysicsAnalysis/JetTagging/FlavorTagDiscriminants/Root/OnnxUtil.cxx

added 23.0 Analysis JetEtmiss analysis-review-required review-pending-level-1 labels

resolved all threads

The system determined that CI tests (with names matching "^CITest_SimulationRun(2|3)(FullSim|Hit).*$") are not needed for this code change. They are not run. This is not an indicator to restart the job.

CI Result SUCCESS (hash 45fbe849)

	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-CC7 73769]

The changes look okay but are extensive, so passing to L2 for comment (this afternoon sometime).

-- James for L1

added review-pending-level-2 label and removed review-pending-level-1 label

Looks good. Karolos (L2)

added review-approved label and removed review-pending-level-2 label