CMSSW code to generate babies (small flat ntuples) from MINIAOD

Code setup

To set up the code and generate a file named baby.root, issue the following commands on lxplus:

cmsrel CMSSW_8_0_26_patch1
cd CMSSW_8_0_26_patch1/src
cmsenv
git cms-init
git cms-merge-topic -u cms-met:METRecipe_8020
git cms-merge-topic -u cms-met:fromCMSSW_8_0_20_postICHEPfilter
git cms-merge-topic -u mverzett:DeepFlavour-from-CMSSW_8_0_21
mkdir RecoBTag/DeepFlavour/data/
cd RecoBTag/DeepFlavour/data/
wget http://home.fnal.gov/~verzetti//DeepFlavour/training/DeepFlavourNoSL.json
cd -
git clone git@github.com:manuelfs/babymaker
cd babymaker
./compile.sh
./scripts/cmsrun.sh <inputfile> <nevents=1000> <outname>

The compile.sh script first compiles the babymaker/bmaker/genfiles folder, which automatically generates the tree structure (see below), and then issues scram b in the babymaker folder.

CMSSW 9 Update

Setup for running on 2017 re-reco data with tag 17Nov2017:

cmsrel CMSSW_9_4_0
cd CMSSW_9_4_0/src
cmsenv
git cms-merge-topic cms-met:METFixEE2017_949_v2
git clone git@github.com:manuelfs/babymaker
cd babymaker
./compile.sh

CMSSW 10

Setup for running on 2018 data and DeepAK8:

cmsrel CMSSW_10_2_1
cd CMSSW_10_2_1/src
cmsenv
# DeepAK8 setup (currently commented out)
    git clone ssh://git@gitlab.cern.ch:7999/DeepAK8/NNKit.git
    # setup mxnet library
    cp NNKit/misc/mxnet_predict.xml $CMSSW_BASE/config/toolbox/$SCRAM_ARCH/tools/selected
    scram setup mxnet_predict
    rm $CMSSW_BASE/external/$SCRAM_ARCH/lib/libmxnet_predict.so
    cp NNKit/misc/lib/libmxnet_predict.so $CMSSW_BASE/external/$SCRAM_ARCH/lib/libmxnet_predict.so
    # compile
    scram b -j16
git clone git@github.com:manuelfs/babymaker
cd babymaker
./compile.sh

Adding new branches

To add new branches to the tree, you first create the new branch in babymaker/variables/full where the type and name are specified. The code in babymaker/bmaker/genfiles/src/generate_baby.cxx automatically generates the files

babymaker/bmaker/interface/baby_base.hh
babymaker/bmaker/interface/baby_full.hh
babymaker/bmaker/src/baby_base.cc
babymaker/bmaker/src/baby_full.cc

which have the c++ code that defines the class baby_full with the tree, all the branches, automatic vector clearing in the baby_full::Fill method, and other useful functions.

The branches are filled in the bmaker_full::analyze method in babymaker/bmaker/plugins/bmaker_full.cc. Functions that define physics quantities, like isolation or electron ID, are defined in babymaker/bmaker/src/*_tools.cc.

Getting an flist

To process entire datasets you need the corresponding flist, or list of individual files in the dataset. A number of these flists are committed to trandbea/flists, and are downloaded with

git clone git@github.com:trandbea/flists

This repository must be placed at ${CMSSW_BASE}/src, not inside babymaker.

To obtain an flist for a new dataset, or a number of datasets, you use the following script

./scripts/get_flist.py -d <dataset_name>
./scripts/get_flist.py -f <file_with_dataset_names>

All flists must be placed in the ${CMSSW_BASE}/src/flists folder.

Submitting jobs to condor in the T3 at UCSB

Log on to one of the compute-0-X machines and set up the code are described above. In the babymaker folder, create an out folder. This is typically a soft link to a place with lots of disk space, such as ln -s /net/cms2/cms2r0/user/out/ out.

Then, add the datasets you want to run over in the wishlist of scripts/sub_cond.py and submit the condor jobs from babymaker folder with

./scripts/sub_cond.py

Post-processing of produced babies

See bmaker/genfiles/README.md for detailed instructions.

Conventions in babymaker

In order to homogenize the code and know what to expect, we try to follow these conventions in the development of babymaker:

Branch names use all lower case letters. If words must be separated, use an underscore, e.g. met_phi
File names use all lower case letters. If words must be separated, use an underscore, e.g. lepton_tools.cc
Function names follow the standard CMSSW convention, that is, first word all lower case, and first letter of subsequent words in upper case. e.g. bmaker_full::writeFatJets
Product names (e.g. "slimmedElectrons") are only defined in bmaker_full::analize or in babymaker/bmaker/python/bmaker_full_cfg.py.
As much as possible, physics definitions go in the corresponding src/*_tools.cc file, e.g. lepton ID goes in src/lepton_tools.cc. They should not be part of plugins/bmaker_full.cc so that when/if we move to having various baby definitions in parallel, for which the code is setup, all babies would use common definitions.