Closed youyingli closed 3 years ago
Hi @edjtscott and @Joe-W-Davies, could you please check if this bug-fixing PR can solve the high/low weight problem mentioned from
hi @youyingli yes for sure, we are going to include it in our next ntuple production to check the weights are returned to normal. We will let you know as soon as it's done
Thanks a lot @youyingli for this PR - @Joe-W-Davies has now tested it and confirmed that the weights now behave sensibly
Thank @edjtscott @Joe-W-Davies for confirmation. The PR is ready to merge.
Because when JobConfig.py includes the pileup information,
mix_2017_25ns_UltraLegacy_PoissonOOTPU_cfi.py mix_2018_25ns_JuneProjectionFull18_PoissonOOTPU_cfi.py
in CMSSW [1] come from the same object which is mix_probFunction_25ns_PoissonOOTPU_cfi.py Therefore, if mix_2017_25ns_UltraLegacy_PoissonOOTPU_cfi.py is first included and then mix_2018_25ns_JuneProjectionFull18_PoissonOOTPU_cfi.py, the information from later will overwrites the previous which is UL17 and vice versa. This results in huge weights (2018 MC pileup are mis-assigned to UL2017 samples). Now, these two are pulled to flashgg and modified to independent formats and bugs are fixed.
[1] https://github.com/cms-sw/cmssw/blob/CMSSW_10_6_X/SimGeneral/MixingModule/python/mix_2017_25ns_UltraLegacy_PoissonOOTPU_cfi.py#L2 https://github.com/cms-sw/cmssw/blob/CMSSW_10_6_X/SimGeneral/MixingModule/python/mix_2018_25ns_JuneProjectionFull18_PoissonOOTPU_cfi.py#L2