Sherpa Tar Creation of Sliced Samples

This topic describes how to create tar balls, which contain full integration and logs, to add to the input for generation and hence shorten the generation-process.

For steps of generation involving Sherpa Tarballs, see: https://twiki.cern.ch/twiki/bin/viewauth/AtlasProtected/SherpaForAtlas#Preparing_input_tarballs_with_sh

Find Sherpa-2.2.8 documentation here: https://sherpa.hepforge.org/doc/SHERPA-MC-2.2.8.html

***************************************************************************************************************************************************************************************

Sherpa Tar Creation

The current version of the Sherpa Tar Creator we've used at Tufts' supercomputer Cluster

..can be found here: https://gitlab.cern.ch/fkaya/sherpaTarCreator/-/tree/master

..implements Sherpa-2.2.8 at /cvmfs/sft.cern.ch/lcg/releases/LCG_88/MCGenerators/sherpa/2.2.8

..uses 32 cores for the "integration" step and 12 cores for the other four steps. When Openloops is necessary (with job options including loops), there is an additional 0th step, namely "getOpenLoops".

• Login and start an interactive node
• Set up Athena-Generation with (recommended for Sherpa_2.2.8) > asetup 21.6.2,AthGeneration
• Pick a new singularity module > module load singularity/3.1.0
• Submit tar creation > python ../sherpaTarCreator/sherpaTarCreator.py MC.File.Of.Your.Choosing --noPrecompile
• Watch progress > watch --interval 30 sacct --job={[step0-jobid]..[step5-jobid]} --format=jobid,jobname,elapsed,ncpus,state

..appears in your runDir, e.g.: mc_13TeV.Sh_228_Zee_EnhLogPtV_BbarMode5.GRID.tar.gz

..can be unzipped by > gunzip -c *.tar.gz | tar xopf -

..will have: 3.integrate.log Process/ Results.db

For an example of these files and the complete logs that are produced during a tar ball creation, see: https://gitlab.cern.ch/fkaya/stitching_tests/-/tree/master/tarBalls%2Fmc_13TeV.Sh_228_Zee_EnhMaxHT_BbarMode5.GRID

***************************************************************************************************************************************************************************************

Job options

Here we'll describe the specific Job Options (e.g., sliced) we've looked at and how we created them.

Original jobId/jobOption is 421304/mc.Sh_228_Zee_EnhLogMaxHTpTV_BbarMode5.py

Link to the original jobOption (JO): https://gitlab.cern.ch/fkaya/stitching_tests/-/blob/master/origJOs/421304/mc.Sh_228_Zee_EnhLogMaxHTpTV_BbarMode5.py

Implemented changes:

Starting with the Original, we've changed the following:

• Reduced the number of events to 100 to increase testing speed
• Reduced NJET to 3 and edited processes so that there are at least 1 jet by redefining "Process 93 93 -> 11 -11 93 93{NJET}" and adding "Cut_Core 1" (so the integration starts with events of type 2_3__j__j__e-__e+__j__QCD, you can check this by looking into runDir/ecm13TeV/3.integrate.log once the integration starts),
• Edited EnhancedVariable for the slice to be only enhancing HT "Enhance_Observable VAR{log10(sqrt(H_T2)-PPerp(p[2])-PPerp(p[3]))}|1|3.7 {3,4,5,6,7}" (instead of the max among HT and pT)
• Defined pT boundaries "PT2NLO 11 -11 500 E_CMS"

Link to the Sliced JO: https://gitlab.cern.ch/fkaya/stitching_tests/-/blob/master/slicingJOs/421364/mc.Sh_228_Zee_EnhMaxHT_BbarMode5.py

*****************************************************************************************************************************************************************************************

Optimization of Cluster jobs

We've tested enhancement boundaries on the HT and boundaries of pT on the tarballs we've generated. Statistics on these tests:

***************************************************************************************************************************************************************************************

-- Colette F I Kaya at FikriyeIdilKaya - 2020-04-18