Apply the lumi normalization (for the given year) in the processor step. It would make sense to do it in the step since we have direct access to the year of the sample (as this info is included in the input json). After this step we no longer have direct access to the year (we have to infer it from the name of the sample axis) so it is more difficult. Furthermore, applying the lumi normalization in this processor step will make the yields in the resulting output histograms easier to directly parse.
As a part of this update, all scripts that take as input the histograms produced by topeft (e.g. plotting scripts, datacard maker, and others) should all be updated to remove the lumi normalization step (to avoid scaling by lumi twice).
Apply the lumi normalization (for the given year) in the processor step. It would make sense to do it in the step since we have direct access to the year of the sample (as this info is included in the input json). After this step we no longer have direct access to the year (we have to infer it from the name of the sample axis) so it is more difficult. Furthermore, applying the lumi normalization in this processor step will make the yields in the resulting output histograms easier to directly parse.
As a part of this update, all scripts that take as input the histograms produced by topeft (e.g. plotting scripts, datacard maker, and others) should all be updated to remove the lumi normalization step (to avoid scaling by lumi twice).