2. Reduced Ntuples
Currently, it has two steps.
Step -1 : Converts the edm to simple ROOT file.
Step -2 : Converts the ROOT file to flat ROOT file. Here, flat means that all the vectors are splitted into separate entries.
2.1. Step -1: Setup
cmsrel CMSSW_12_0_1
cd CMSSW_12_0_1/src
cmsenv
git cms-merge-topic Sam-Harper:L1NtupleFWLiteFixes_1130pre4
# git clone ssh://git@gitlab.cern.ch:7999/sharper/HLTAnalyserPy.git Analysis/HLTAnalyserPy
git clone -b RegNtupleRun3 ssh://git@gitlab.cern.ch:7999/rasharma/HLTAnalyserPy.git Analysis/HLTAnalyserPy
scramv1 b -j 8
Command to run for small statistics:
python3 Analysis/HLTAnalyserPy/test/makeRun3Ntup.py <input>*.root -o test.root -r 1000
Command to run over large statistics (Using 8 cores):
# Using multithread
python3 Analysis/HLTAnalyserPy/test/runMultiThreaded.py -o output.root <input>*.root --cmd "python3 Analysis/HLTAnalyserPy/test/makeRun3Ntup.py -r 50000" --hadd
2.2. Step - 2: Get Flat ROOT File
Simple script is placed inside the EgRegresTrainerLegacy, here. One can use it as
root -l -b -q GetFlatNtuple.C\(\"Input.root\",\"Input_Flat.root\"\)
Note
Before moving to the next step, which is the regression training, we should have two flat root files. They are
ideal flat root file
read flat root file