TWiki> ATLAS Web>HiggsBb (revision 26) Edit Raw edit Attach Print version

Introduction

Documenting the setup of Higgsbb analysis code used by cern.

An overview of the software framework can be found here.

The CERN twiki page for Diboson studies may also be useful.

The latest CERN ntuples may be useful for testing purposes.

(Note: If at any stage when setting up an environment or building code you encounter a problem before trying anything else first do make clean on anything you have done make on then use a clean shell and retry.)

Running Mode

The code can be run in two different modes. Locally or using PROOF(Parallel ROOT Facility). PROOF allows analysis of a large number of ROOT files in parallel using multiple machines or processor cores.

Running Locally

Code Checkout

The latest version of the code can be checked out from the correct repository by doing:

mkdir myVH_research
cd myVH_research
svn co svn+ssh://svn.cern.ch/reps/atlasusr/mbellomo/ElectroweakBosons/trunk

However it is probably best to use the latest tagged version of the code as this should (but wont necessarily) be stable. This can be checked out by doing

mkdir myVH_research
cd myVH_research
svn co svn+ssh://svn.cern.ch/reps/atlasusr/mbellomo/ElectroweakBosons/tags ElectroweakBosons-xx-xx-xx 

Where xx-xx-xx is the tag number.

Note: you may want to set up kerberos first.(I did not require the GSSAPITrustDNS yes line).

Compiling the Code

(Note: for more complete instructions see the README file in the trunk subdirectory)

(1) To setup the environment:

cd myVH_research/trunk
source scripts/setup_lxplus.sh

This will setup the required environment variables, the Gnu C++ compiler and ROOTCORE.

(2) To download and compile and extra software required do

(Note: If the extra code fails to compile due to the MET package you need to remove the code and run the 2012 versions of all the scripts)

./scripts/get_allcode.sh
cd SFrame
source setup.sh
cd ..

(3) Next compile the software by doing

make

(You can also clean up by doing make clean and make distclean)

(4) For the purposes of analysing H->bb decays the code is stored in the AnalysisWZorHbb/ directory.We now need to compile this code before we can do the analysis.

cd AnalysisWZorHbb
make

Configuring and Running

Now we can run our analysis by doing sframe_main config/WZorHbb_config_mc_nu_2011.xml for example. The .xml file is where the input files and settings are all defined(e.g. what corrections to apply to the data) and can be modified or written to suit the needs of the analysis. For example to do a H->bb analysis involving zero leptons that makes flat ntuples you need the following lines.

 <UserConfig>
                &setup;
                <Tool Name="BaselineZeroLepton"    Class="WZorHbb">
                        &tool_base;
                                <Item Name="runNominal" Value = "True"/>
                                <Item Name="ntupleName" Value = "Ntuple"/>
                </Tool>
    </UserConfig>

The "BaselineZeroLepton" option is what sets the analysis type. The string is parsed to determine whether we are doing a 0, 1 or 2 lepton analysis and also if we want to do an electron, muon or nominal analysis. (See line 250ish in WZorHbb.cxx for the code that performs the selection)

(Note: every time you open a shell to run the code you need to do steps (1), (3) and (4) from the compilation instructions again. i.e. setup the environment and make the code. It is also a good idea to run make clean before you recompile anything.)

(Note: It is advisable to use BaselineOneLepton for both the 1 lepton and 2 lepton analysis then diffirentiate between the two in the config file that is passed to RunHistmaker.C)

In order to do a full analysis with the whole CERN framework this code must produce three files. One for montecarlo, one for electrons and one for muons. To do this three separate "cycles" must be defined in the .xml file. A cycle is defined as follows


<Cycle Name="AnalysisManager" RunMode="LOCAL" ProofServer="lite://" ProofWorkDir="/afs/cern.ch/user/p/pmullen/tag-tmp/" ProofNodes="-1" UseTreeCache="True" TreeCacheSize="30000000" TreeCacheLearnEntries="10" OutputDirectory="/afs/cern.ch/user/p/pmullen/tag-tmp/" PostFix="" TargetLumi="1.0">

    <InputData Lumi="1.0" NEventsMax="-1" Type="mc12_8TeV" Version="OneLepton.Nominal" Cacheable="True" SkipValid="True">

        &pauls_mc;
       <InputTree Name="physics" />
                 <OutputTree Name="Ntuple" />
    </InputData>

<UserConfig>


                &setup;
                <Item Name="CorrectFatJet"  Value="False"/>

                <Tool Name="BaselineOneLepton"    Class="WZorHbb">
                        &tool_base;
                                <Item Name="runNominal" Value = "True"/><!--true-->
                                <Item Name="ntupleName" Value = "Ntuple"/>

                </Tool>
    </UserConfig>
  </Cycle>

Where &pauls_mc is a file with paths to the montecarlo files being input. For contrast another cycle configuration is shown below; this time for an electron analysis in data.

(Note: The first cycle must be closed with the flag before opening another cycle)

  <Cycle Name="AnalysisManager" RunMode="LOCAL" ProofServer="lite://" ProofWorkDir="/afs/cern.ch/user/p/pmullen/tag-tmp/" ProofNodes="-1" UseTreeCache="True" TreeCacheSize="30000000" TreeCacheLearnEntries="10" OutputDirectory="/afs/cern.ch/user/p/pmullen/tag-tmp/" PostFix="" TargetLumi="1.0">


    <InputData Lumi="1.0" NEventsMax="-1" Type="data12_8TeV" Version="OneLepton.Electron" Cacheable="True" SkipValid="True">    

        &pauls_data;
       <InputTree Name="physics" />
                 <OutputTree Name="Ntuple" />
    </InputData>
    <UserConfig>

                &setup;
                <Item Name="CorrectFatJet"  Value="False"/>

                <Tool Name="OneLepton"    Class="WZorHbb">
                        &tool_base;
                                <Item Name="runNominal" Value = "True"/>
                                <Item Name="ntupleName" Value = "Ntuple"/>
                </Tool>


    </UserConfig>

  </Cycle>


The config file and the file containing the paths to the montecarlo files are attached for reference.

Producing Plots

The output from running sframe is stored in .root files as common ntuples. To produce plots from these output files we need to use HistMaker by doing

cd macros
make

to make the required code. To run the code we use the command

./RunHistMaker  <path to config file> <runMJ> <path to files output by sframe> <output directory>      -- explain better?

e.g.

./RunHistMaker  configs/zerolepton.config false $HOME/out/AnalysisManager.mc11_7TeV.ZeroLepton.Nominal.root ./


This will output a .root file containing histograms.

If RunHistMaker crashes saying it did not find the cross-section(xsec) for a given number(correlating to a data set) you may need to edit the function InitCrossSections2011() in the file macros/src/process.cpp. You will need to add the following code


allXSecs.push_back(xsection(dataset, cross-section, k factor, 1, "sample name"));  -- find out what 1 is.

e.g.

allXSecs.push_back(xsection(109351, 4.0638E-05, 1, 1,"ZHnunu"));

Atlas twiki pages can be used to find the cross-section for 7TeV data, 8 TeV data and for general datasets.

For a 2012 event InitCrossSection2012() pulls the information from a file called configs/2012_cross_sections.txt which can be edited to add the information for your particular dataset

Adding Extra Plots

Stacking Plots

Stacked plots can be produced using the RunPlotMaker.C program.(This uses the plotter.cpp class) The current implementation of the program expects input files to follow a naming convention that contains "OneLepton" in the root file name. This need to be changed depending on what you had in your .xml file. It also has hard coded links to a home directory that need to be changed. Choosing what systematics etc. you want to run is done with command line flags. My current version of the edited RunPlotMaker.C can be found here. It has been edited so that you can give it the path to a folder containing your files as a command line argument. It can be run as follows

./RunPlotMaker -Data12 -llbb -NominalLimitFile -Indir $HOME/out/

It expects to find 6 files in the $HOME/out/ folder.

AnalysisManager.data12_8TeV.OneLepton.Electron.Hists.MJ.root
AnalysisManager.data12_8TeV.OneLepton.Electron.Hists.root
AnalysisManager.data12_8TeV.OneLepton.Muon.Hists.MJ.root
AnalysisManager.data12_8TeV.OneLepton.Muon.Hists.root
AnalysisManager.mc12_8TeV.OneLepton.Nominal.Hists.MJ.root
AnalysisManager.mc12_8TeV.OneLepton.Nominal.Hists.root

Note the "OneLepton" in all the file names. This is because I have yet to change the hard coded file name format in the script. The "MJ" files(Multijet) were produced from the RunHistMaker program by changing the RunMJ parameter to True. If you do not want to have to rename your files before using this program then simply edit your sframe .xml config file as follows

     

<InputData Lumi="1.0" NEventsMax="-1" Type="mc12_8TeV" Version="OneLepton.Nominal" Cacheable="True" SkipValid="True">

        <input data files here>
       <InputTree Name="physics" />
                 <OutputTree Name="Ntuple" />
    </InputData>

For the data files change the Type to "data12_8TeV" and the Version to "OneLepton.Muon" or "OneLepton.Electron". This does not change the analysis type, only the output file name.

To change the Analysis type you change the Tool Name="BaselineTwoLepton" to BaselineTwoLeptonElectron or BaselineTwoLeptonMuon

Adding Extra Stacked Plots

Running Using PROOF on Demand(PoD)

Proof is a framework for running many root analyses in parallel. It can be run locally using PROOF-lite which will utilise multiple cores on one machine or it can be run on many machines usind the PoD framework.

Installing and Building

To Install PROOF on Demand first we must get and unpack the code.

cd $HOME
wget http://pod.gsi.de/releases/pod/3.10/PoD-3.10-Source.tar.gz
tar -xzvf PoD-3.10-Source.tar.gz

To build the code do

cd $HOME/PoD-3.10-Source
mkdir build
cd build
cmake -C ../BuildSetup.cmake ..
make
make install
cd
cd myVH_research/ElectroweakBosons/trunk
source scripts/setup_pod_lxplus.sh
cd
cd PoD/3.10/
source PoD_env.sh

To change the working directory PoD uses along with many other settings just edit $HOME/.PoD/PoD.cfg

(Note: For those using an afs directory you need to change your [server] work_dir parameter to something like /tmp/$USER/PoD)

Now create a file called $HOME/.PoD/user_worker_env.sh and put the line source /afs/cern.ch/sw/lcg/contrib/gcc/4.3.5/x86_64-slc5-gcc43-opt/setup.sh in it

Then create a file called $HOME/.PoD/user_xpd.cf and put the line xpd.rootd allow in it.

DEPRECATED?----------------------------------------------------------------------------------------------------------

Now create a file called $HOME/.PoD/user_worker_env.sh and put the following settings in it


#! /usr/bin/env bash
echo "Setting user environment for workers ..."
export LD_LIBRARY_PATH=/afs/cern.ch/sw/lcg/external/qt/4.4.2/x86_64-slc5-gcc43-opt/lib:\
/afs/cern.ch/sw/lcg/external/Boost/1.44.0_python2.6/x86_64-slc5-gcc43-opt//lib:\
/afs/cern.ch/sw/lcg/app/releases/ROOT/5.30.01/x86_64-slc5-gcc43-opt/root/lib:\
/afs/cern.ch/sw/lcg/contrib/gcc/4.3.5/x86_64-slc5-gcc34-opt/lib64:\
/afs/cern.ch/sw/lcg/contrib/mpfr/2.3.1/x86_64-slc5-gcc34-opt/lib:\
/afs/cern.ch/sw/lcg/contrib/gmp/4.2.2/x86_64-slc5-gcc34-opt/lib
echo "LD_LIBRARY_PATH=$LD_LIBRARY_PATH"

END---------------------------------------------------------------------------------------------------------------------------------------

Using PROOF

We now have PROOF installed but before doing any analysis we must first setup the correct environment. It is advised that after installing PROOF you start from a clean shell then do

cd myVH_research/trunk
source scripts/setup_lxplus.sh
source scripts/setup_pod_lxplus.sh
make
cd AnalysisWZorHbb
make

(Note: It is advisable to delete your ~/.proof and /tmp/$USER directories before starting proof)

To start the proof server we can do pod-server start. You can now request worker nodes. To do this we use the pod-submit command. For example to request 20 worker nodes we would do

pod-submit -r lsf -q 1nh -n 20

Where 1nh is the name of a que we are submitting the jobs to and lsf is the name of the resource management system.

Now that we have set up our environment and requested worker nodes we can use the nodes for our analysis. To run analysis we use the same comand as before; sframe_main config/proof_WZorHbb_config_mc_nu_2011.xml.

(Note: It is advisable to use nohup sframe_main config/proof_WZorHbb_config_mc_nu_2011.xml &> output.txt & so that terminal crashes will not kill your job)

However our .xml file will look different from our non-PROOF analysis. We must change the RunMode option to be RunMode="PROOF" and the ProofServer option must also be changed. It must look like ProofServer="username@host:portnumber" where the port number, host and username are those output from the pod-server start command or the pod-info -c. For example ProofServer="pmullen@lxplus438.cern.ch:21002"

Adding D3PD variables to our Algorithm

In order to use D3PD variables that are not already used by the CERN code you must first find the details of the variable. To do this it is easiest to load the root file in CINT (root -l myfile.root) then use the MakeClass function of root to produce a header file that could be used to read the tree. This can be done on the CINT command line by doing treename->MakeClass("MyClassName"). This will write out a MyClassName.C and a MyClassName.h file. The details of how to read the variable you want can then be found in these files.

Once you have the details of the variable we can now add it to the AnalysisBase/include/EventBase.h and the AnalysisBase/include/Event.h file so that any code that inherits from this code is aware of the variable(the AnalysisWZorHbb code inherits from this as does all the code that comes from the CERN SVN package)

Now that we have done this we can add our variable to our function. In the case of the AnalysisWZorHbb code we would go to /AnalysisWZorHbb/src/WZorHbb.cxx and add ConnectVariable("MyD3PDVariableName") to the BeginInputFile function.

For instance for the jet_AntiKt4TopoEM_WIDTH variable we would find in MyClassName.h the line vector   *jet_AntiKt4TopoEM_WIDTH;. We would then go to AnalysisBase/include/EventBase.h and add this line there. We would then go to /AnalysisWZorHbb/src/WZorHbb.cxx and add ConnectVariable("jet_WIDTH"). Note that we have removed the name of the algorithm from the variable name. This is the convention used. The algorithm name is added by the ConnectVariable function. We must also add vector *jet_WIDTH to AnalysisBase/include/Event.h .

Adding the Variable to The Output Tree

For the variable to be written to the TTree in the SFrame section of the code we must create an output variable. To do this first add a variable of the correct type to the Algorithms header file. Now go to the Algorithms source file and use DeclareVariable in the BeginInputData() function to declare a branch in the output tree. Make sure to also clear() the variable in the ResetNtupleVars() function. You can now fill this variable in the fillNTuple() function probably using push_back.

As an example lets say we want to output the jet_WIDTH variable from the previous section. We would first add std::vector pauls_jet_WIDTH; in the header file. In the source file we would go to the BeginInputData() function and add DeclareVariable(pauls_jet_WIDTH, "pauls_jet_WIDTH", treename); where pauls_jet_WIDTH is the name of the branch in our output tree and treename is our output tree; Also add pauls_jet_WIDTH.clear(); in the fillNTuple() function. To fill this variable we would then add pauls_jet_WIDTH.push_back(ev.jet_WIDTH->at(i)); to the fillNTuple() function.

Topic attachments
I Attachment History Action Size Date Who Comment
Unknown file formatC RunPlotMaker.C r2 r1 manage 35.1 K 2013-02-22 - 12:32 PaulMullen plotting script
XMLxml llbb_local.xml r1 manage 4.2 K 2013-02-22 - 12:50 PaulMullen Path to montecarlo files
XMLxml paul_test_lepton_cuts.xml r1 manage 4.5 K 2013-02-22 - 12:50 PaulMullen SFrame configuration file
Edit | Attach | Print version | History: r32 | r28 < r27 < r26 < r25 | Backlinks | Raw View | Raw edit | More topic actions...
Topic revision: r26 - 2013-11-25 - PaulMullen
 
This site is powered by the TWiki collaboration platform Powered by PerlCopyright © 2008-2020 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback