Toolbox for resonant HH analysis
This repository consists of scripts interfacing the CommonTools and PlotIt facilities. It does not require any compilation, just follow this instructions below to setup everything.
We use Keras and Tensorflow for Neural Network training / evaluation, which is not available in CMSSW. When first installing this repository, you must register an already pre-built version of Keras and Tensorflow by executing the add_keras_to_cmssw.sh
script at the root of the repository. Note: you must first be in a CMSSW environment by doing a cmsenv
.
cms_env
cmsenv
with, in the .bashrc,
alias cms_env="module purge; module load grid/grid_environment_sl6; module load crab/crab3; module load cms/cmssw;"
.
This step is done in histFactory_hh
. You need a python script to generate the plots (such as generatePlots.py
based on basePlotter.py
). The important is that this code defines a list called plots
. To launch the TH1(2) production, use
python launchHistFactory.py -o OUTPUTFOLDER -s -r
.
The -s
option actually submit the jobs on condor, use it first without to check that there is no error with source OUTPUTFOLDER/condor/input/condor_1.sh
. The -r
option removes OUTPUTFOLDER
if it exists. NB : to avoid risky command, it first prompts the command, you need to confirm by typing "enter".
This step is done in plotIt_hh
. To launch the plots :
./plotIt.sh [SUFFIX]
will create a folder named plots_all_DATE_SUFFIX
with all the stacked plots inside, SUFFIX being optional.
Here are the key files you have to modify :
centralConfig.yml
: to specify the directory where the rootfiles are.
listHistos.py
: you can modify the rootFile at the top and run python listHistos.py
to generate the .yml file with axis labels, blinded range, etc.
MCFiles.yml
and DataFiles.yml
to specify the samples you want to run on. Usually you just need to replace the old tag by the new one (e.g. in vim :%s/v2.0.3+7415_HHAnalysis_2016-01-30.v3/v2.0.4+7415_HHAnalysis_2016-02-14.v0/
).
groups.yml
to specify the colors, the names and which samples should appear merged in the legend.