/AZH_analysis

Primary LanguagePython

Princeton A/ZH Framework

Framework for the A->ZH->lltt and ZH->lltt analyses.

Setup

This analysis framework is configured to work with the cmslpc-sl7 cluster, which you can ssh into with ssh username>@cmslpc-sl7.fnal.gov. For full details about setting up your computing environment, please see the USCMS Software Setup Guide. First, source the environment setup script:

source /cvmfs/cms.cern.ch/cmsset_default.sh (for bash, .csh for cshell)

You'll need to set up the analysis in the work area of a CMSSW release:

cmsrel CMSSW_10_2_9
cd CMSSW_10_2_9/src
cmsenv

pip

Each script has various dependencies, which you may or may not have installed in your Python3 environment. For example, you may need to install yaml like so:

pip install --user pyyaml

conda

I find it helpful to keep my software organized in Conda environments. At Fermilab's recent ML Hands-on Tutorial Session, they provide the following recipe for installing conda on the LPC:

wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh -O $HOME/nobackup/miniconda3.sh
bash $HOME/nobackup/miniconda3.sh -b -f -u -p $HOME/nobackup/miniconda3
source $HOME/nobackup/miniconda3/etc/profile.d/conda.sh

You can then create a conda virtual environment and install relevant packages within it, for example:

conda create --name AZH_env python=3.6
conda install -c conda-forge root # install root in conda env
conda install matplotlib          # install other packages via conda
conda install pip                 # add pip to the conda env

# pip install within your conda env:
/uscms/home/<username>/nobackup/miniconda3/envs/AZH/bin/pip install coffea

Note that you should not source cmsset_default.sh if you're working in a conda environment - minimally, this breaks root.

Job Submission

Jobs are submitted to the Condor cluster following examples listed in the USCMS Batch Systems Guide and the Submitting Multiple Jobs Using HTCondor page.

Pileup Re-weighting

Utilities for Accessing Pileup Information for Data

Columnar Analysis

This framework uses a number of tools developed for columnar analysis, which views particle physics datasets as matrices with columns corresponding to variables of interest (as opposed to event-by-event conceptualization). Some key resources used to develop this framework include: