Skip to content

postProcGuide.md

Latest commit

 

History

History
176 lines (123 loc) · 11.8 KB

postProcGuide.md

File metadata and controls

176 lines (123 loc) · 11.8 KB

Postprocessing guide for LatinoAnalysis NanoGardener

This guide is divided in two sections. In the first one the reader is instructed on how to process centrally produced nanoAOD, enabling the processing of virtually any sample for your analysis (in particular this is extremely useful for most common background processes such as DY, multijet QCD, top induced processes etc.. ). Bear in mind that the Latinos group already processed plenty of samples, so before starting the postprocessing of a sample, check the following files and see wether your sample has been already processed:

The second part is devoted to the processing of nanoAODs from a private production. In the real life, you might want to study a graviton production decaying into an Higgs pair decaying into four muons. You quickly realize no one ever did such search (because it is quite dumb) and therefore there are no centrally produced nanoAOD. You then create your gridpack, run it through the whole simulation up to nanoAODv7 level and then you might want to further postprocess the sample to use the Latinos framework for your study.

Process centrally produced nanoAOD

Suppose you started a new and very fancy analysis. You realize that somehow one of the needed background is not available in the Latinos central folder. First thing you need to check wether centrally produced nanoAOD do exist (which are "vanilla nanoAOD", before the Latino post-processing). To do so you might want to consult the CMS Data Aggregation System: https://cmsweb.cern.ch/das/request?view=list&limit=50&instance=prod%2Fglobal&input=dataset%3D%2FDYJetsToLL_0J*%2F*NanoAODv7*102X*%2FNANOAODSIM

Pay attention to the syntax you use to query datasets. The first entry DYJetsToLL_0J* is just the dataset name, followed by the MC Tune TuneCP5 for underlying events, the COM energy 13TeV, the hard process MC generator amcatnlo, the parton-jet marging algorithm FXFX and the parton shower pythia8. The second entry is the central production camapign RunIIAutumn18NanoAODv7-Nano02Apr2020_102X which tells you the date of the campaign, the nanoAOD version for the campaign NanoAODv7 and the CMS version used for the processing 102X. It is important that the samples you select are processed with version 7 of the nanoAOD NanoAODv7 as the Latino Frameworks only works with this version. The last line is the global tag and version (detector geometry etc etc) used for the campaign upgrade2018_realistic_v21-v1. Lastly, and this is important, the dataset type you want in our case it is mandatory the NANOAODSIM step (there are many others, slide 26 here https://conference.ippp.dur.ac.uk/event/704/attachments/3462/3813/CMS_data_mc_IPPP.pdf ).

Once you find the dataset you want to process with Latinos framework e.g. /DYJetsToLL_0J_TuneCP5_13TeV-amcatnloFXFX-pythia8/RunIIAutumn18NanoAODv7-Nano02Apr2020_102X_upgrade2018_realistic_v21-v1/NANOAODSIM you have to edit some Latinos files in order for xrdcp to take this root filles distributed over the various Tier2, issue the processing, and copy the output on your personal eos (by default it will try to copy the output to /eos/cms/store/group/phys_higgs/cmshww/amassiro/HWWNano/ which is a special directory with limited write access.

In Sites_cfg.py line 16 (https://github.com/latinos/LatinoAnalysis/blob/76e7c4b93aa5f056c92440d4e8d24e7de749c8fe/NanoGardener/python/framework/Sites_cfg.py#L16) modify the dict entry as follows:

  'cern' : {
              'lsCmd'       : 'ls' ,
              'mkDir'       : True ,
              'xrootdPath'  : 'root://eosuser.cern.ch/' ,
            #  'treeBaseDir' : '/eos/cms/store/group/phys_higgs/cmshww/amassiro/HWWNano/' ,
              'treeBaseDir' : '/eos/user/g/gboldrin/nanoLatino/' ,
              'treeBaseDirOut' : '/eos/user/g/gboldrin/nanoLatino/' ,
              'batchQueues' : ['8nh','1nd','2nd','1nw'],
              'slc_ver'     : 7
           } ,

where the entries treeBaseDirand treeBaseDirOut should point to an existing folder on your personal eos. Secondly uncomment the following line https://github.com/latinos/LatinoAnalysis/blob/76e7c4b93aa5f056c92440d4e8d24e7de749c8fe/NanoGardener/python/framework/PostProcMaker.py#L32

Now you just need to add your sample to the samples configuration (https://github.com/latinos/LatinoAnalysis/blob/76e7c4b93aa5f056c92440d4e8d24e7de749c8fe/NanoGardener/python/framework/samples/Autumn18_102X_nAODv7.py). The order is not important. Just add an entry where the key DYJetsToLL_0J in this example is arbitrary:

Samples['DYJetsToLL_0J'] = {'nanoAOD' : '/DYJetsToLL_0J_TuneCP5_13TeV-amcatnloFXFX-pythia8/RunIIAutumn18NanoAODv7-Nano02Apr2020_102X_upgrade2018_realistic_v21-v1/NANOAODSIM'}

Now you can run the first step of the post-processing as follows:

mkPostProc.py -p Autumn18_102X_nAODv7_Full2018v7 -b -s MCl1loose2018v7 -Q tomorrow -T DYJetsToLL_0J

Where no -i step should be specified as we are starting from scratch. In order to further process this sample after this step is finished some changes are needed because we would want to take the output of the MCl1loose2018v7 step which is now on our eos space. In order to do such please read the next chapter.

Process privately produced nanoAOD

Step Zero

Try to run eos ls /eos/user/INITIAL-OF-YOUR-NAME/YOUR-ACCOUNT/. If it raises an exeption you should add a simple line in your .bashrc or .zshrc: export EOS_MGM_URL=root://eoshome-INITIAL-OF-YOUR-NAME.cern.ch and then source your bashrc/zshrc. This sets the right redirector of your EOS. If also with this fix you get an error try with the following: root://eosuser.cern.ch. Now that you know what your redirector is, you can use eos and xrdcp without any trouble.

In order to produce 102X nanoAOD, one should use 10_2_X or newer release

cmsrel CMSSW_10_6_4
cd CMSSW_10_6_4/src
cmsenv
git clone --branch 13TeV git@github.com:latinos/setup.git LatinosSetup
source LatinosSetup/SetupShapeOnly.sh
scram b

Copy the file LatinoAnalysis/Tools/python/userConfig_TEMPLATE.py to LatinoAnalysis/Tools/python/userConfig.py and edit it to reflect your local paths.

Now we will be editing some files in LatinoAnalysis/NanoGardener/python/framework

First step

Change the Sites_cfg.py. The content with key 'cern' is the one you will be using and should point to your EOS output folder (i.e. where your post processed ntuples will be saved). Specify the EOS url you used in step zero as xrootdPath, and the path of your output folder as treeBaseDir. You should end up with something like:

'xrootdPath': 'root://eoshome-g.cern.ch/',
'treeBaseDir': '/eos/user/g/gpizzati/nanoAOD/postProc',

Second step

To make the framework input root files in your eos (I suggest to put your private nanoAOD in your eos folder), then you need to change PostProcMaker.py:

We want the framework to know the right redirector to use, it should be the same url you used above. Change this line to something like: self._aaaXrootd = 'root://eoshome-g.cern.ch/'

Third step

Add your samples in the file samples/Autumn18_102X_nAODv7.py. Create a new key-value pair in the Samples dictionary and specify in srmPrefix your eos url and in paths the path of your input nanoAOD (i.e. the ones you generated in FULL-SIM)

Here's an example:

Samples['hmn_m50']={'srmPrefix':'root://eoshome-g.cern.ch/',
                    'paths' :['/eos/user/g/gpizzati/nanoAOD/Zjj_SM_5f']}

Fourth

Add your samples cross sections in samples/samplesCrossSections2018.py. You should add a key-value pair in the samples dictionary. e.g. : samples['EWKZ_2Jets_No_Higgs'].extend( ['xsec=1.0', 'kfact=1.000', 'ref=Z' ]) The xs used can be the one given by MG at parton level.

If you used D6EFTStudies, you will find it in the file RESULTS_FOLDER/read_03_input.cfg after the postProcess step.

Finally you can set voms proxy with: voms-proxy-init -voms cms -valid 192:00, and run the commands below.

Remember to run grid before all of the commands below if your proxy expires

Final commands

  1. mkPostProc.py -p Autumn18_102X_nAODv7_Full2018v7 -s MCl1loose2018v7 -b -Q workday -T YOUR_SAMPLE_NAME

  2. mkPostProc.py -p Autumn18_102X_nAODv7_Full2018v7 -s MCCorr2018v7 -i MCl1loose2018v7 -b -Q workday -T YOUR_SAMPLE_NAME

  3. mkPostProc.py -p Autumn18_102X_nAODv7_Full2018v7 -s l2loose -i MCl1loose2018v7__MCCorr2018v7 -b -Q workday -T YOUR_SAMPLE_NAME

If the last step isn't working, please edit the following files in LatinoAnalysis/NanoGardener/python/data/DYSFmva/2018_alt/:

TMVAClassification_PyKeras_2018_0j.weights.xml
TMVAClassification_PyKeras_2018_1j.weights.xml
TMVAClassification_PyKeras_2018_2j.weights.xml
TMVAClassification_PyKeras_2018_VBF.weights.xml
TMVAClassification_PyKeras_2018_VH.weights.xml

A simple sed should fix the problem (supposing you're in CMSSW/src)

array=( 0j 1j 2j VBF VH )
for i in "${array[@]}"
do
	sed -i 's@<Option name="FilenameTrainedModel" modified="No">.*@<Option name="FilenameTrainedModel" modified="No">'`pwd`'/LatinoAnalysis/NanoGardener/python/data/DYSFmva/2018_alt/TrainedModel_PyKeras_2018_'${i}'.h5</Option>@g' LatinoAnalysis/NanoGardener/python/data/DYSFmva/2018_alt/TMVAClassification_PyKeras_2018_${i}.weights.xml
done
  1. mkPostProc.py -p Autumn18_102X_nAODv7_Full2018v7 -s l2tightOR2018v7 -i MCl1loose2018v7__MCCorr2018v7__l2loose -b -Q workday -T YOUR_SAMPLE_NAME