Skip to content

Draft: check if adding soft electron variables in r22 training-dataset-dumper

Junwen Xiong requested to merge jxiong/training-dataset-dumper:xjw-r22-merge into r22

This MR tries to add 32 soft electron variables for 3 phf cuts (~80%, ~90%, 100% efficiencies in selecting true soft electrons) to the existing training-dataset-dumper (TDD) framework.

The code is tested locally and in grid submission, processing the default training DAOD (mc20_13TeV.410470.PhPy8EG_A14_ttbar_hdamp258p75_nonallhad.deriv.DAOD_PHYSVAL.e6337_s3681_r13167_p4931 as suggested in the ftag documentation and TDD repo).

The default DAOD does not contain electron ID DNN scores as the PHYSVAL imports ElectronsCPDetailedContent but not ElectronsCPContent. And, ElectronsCPContent contains information about the DNN scores. Therefore, we use calculateMultipleOutputs in AsgElectronSelectorTool to calculate the DNN scores. In addition, EventContext from GaudiKernel is needed for the calculation. Thus, to add electron variables, TDD can only be compiled with Athena instead of Analysisbase.

Most of the electron variables are added following existing code examples:

This MR is related to the ftag-doc issue: atlas-flavor-tagging-tools/algorithms/ftag-docs#17 (closed)

(@mguth for information)

Edited by Junwen Xiong

Merge request reports