Running the training-dataset-dumper on a release 21 FTAG sample

Hi,
I’m currently trying to dump our newly produced R21 Z + jets samples with the dumper (r22 branch). I’m using the ca-dump-upgrade script to do so. I set up Athena (via `source training-dataset-dumper/setup-athena.sh) and have also done the build steps. Unfortunately I get the following error when running over the file:

PoorMansIpAugmenterAlg                              FATAL  Standard std::exception is caught 
PoorMansIpAugmenterAlg                              ERROR SG::ExcBadAuxVar: Attempt to retrieve nonexistent aux data item `::definingParametersCovMatrixDiag' (15).
AthAlgSeq                                           FATAL  Standard std::exception is caught 
AthAlgSeq                                           ERROR SG::ExcBadAuxVar: Attempt to retrieve nonexistent aux data item `::definingParametersCovMatrixDiag' (15).
AthAllAlgSeq                                        FATAL  Standard std::exception is caught 
AthAllAlgSeq                                        ERROR SG::ExcBadAuxVar: Attempt to retrieve nonexistent aux data item `::definingParametersCovMatrixDiag' (15).
AthAlgEvtSeq                                        FATAL  Standard std::exception is caught 
AthAlgEvtSeq                                        ERROR SG::ExcBadAuxVar: Attempt to retrieve nonexistent aux data item `::definingParametersCovMatrixDiag' (15).
AthMasterSeq                                        FATAL  Standard std::exception is caught 
AthMasterSeq                                        ERROR SG::ExcBadAuxVar: Attempt to retrieve nonexistent aux data item `::definingParametersCovMatrixDiag' (15).

I’m not sure where this is coming from tbh. I’m currently using the upgrade.json config which is provided in the dumper.

Somebody knows what the problem is or how to fix it?

Can you see if that variable exists in the output from checkFile <(D)xAOD>? Also consider specifying what format this is running on, in case we find out that this is specific to one format in particular.

The format is FTAG1. The Dataset is mc16_13TeV.366010.Sh_221_NN30NNLO_Znunu_P TV70_100_BFilter.deriv.DAOD_FTAG1.e6695_e5984_s3126_r10201_r10210_p4870.

When I use checkFile <File> | grep definingParametersCovMatrixDiag, it is not listed there.

Ok I just tried the same with the pure AOD (mc16_13TeV.366010.Sh_221_NN30NNLO_Znunu_PTV70_100_BFilter.merge.AOD.e6695_e5984_s3126_r10201_r10210) and this seems to work. At least the basic config.

EDIT: It is working for the provided upgrade.json (which uses EMTopo) but not for PFlow.

OK, so it sounds like FTAG1 is just missing the diagonal elements of the covariance matrix, right? I’m a bit surprised because I thought people had been running e.g. retagging in the performance framework on these samples, and retagging should also require a covariance matrix.

I’m not sure why it is not working. With the release 21 branch of the training-dataset-dumper, I can run the FTAG1 samples without a problem.

Are we asking for the covariance matrix in the release 21 software? In some cases in release 22 we ask for them when we ask for e.g. q/p, phi, or theta uncertainty.

I’m not sure tbh. The problem already starts when I’m trying switch the JetCollection to PFlow jets…

FTAG1 also probably already has the btagIp_* variables saved, so the algorithm that is throwing this exception can be disabled.