Skip to content

Latest commit

 

History

History
206 lines (140 loc) · 12.3 KB

README.md

File metadata and controls

206 lines (140 loc) · 12.3 KB

PFNano

This is a NanoAOD framework for advanced developments of jet algorithms.

The repository consists of multiple branches which are each dedicated to specific releases of CMSSW. If you came here to run over Run3 samples, please checkout the most up-to-date 12_4_8 branch (e.g. from the dropdown menu above). The master branch you are viewing right now is optimized to run over Run2 samples, using the 106X release cycle.

The current full content of this development branch can be seen here and the size here. In this version, PFcandidates can be saved for AK4 only, AK8 only, or all the PF candidates. More below. This format can be used with fastjet directly.

Recipe

THIS IS A DEVELOPMENT BRANCH

For UL 2016, 2017 and 2018 data and MC NanoAODv8 according to the XPOG and PPD recommendations:

cmsrel  CMSSW_10_6_20 # in principle not a constraint
cd  CMSSW_10_6_20/src
cmsenv
git cms-rebase-topic andrzejnovak:614nosort
git clone https://github.com/cms-jet/PFNano.git PhysicsTools/PFNano
scram b -j 10
cd PhysicsTools/PFNano/test

Note: When running over a new dataset you should check with the nanoAOD workbook twiki to see if the era modifiers in the CRAB configuration files are correct. The jet correction versions are taken from the global tag.

Local Usage:

There are python config files ready to run in PhysicsTools/PFNano/test/ for the UL campaign of nanoAODv8, named nano106Xv8_on_mini106X_201*_data_NANO.py. Notice that the current version can create different types of files depending on the PF candidates content.

New since Pull Request #39: Examples to include or exclude the input features for the DeepJet tagger are given in nano106Xv8_on_mini106X_2017_mc_NANO.py. Now the list of options that are currently implemented inside pfnano_cff.py (e.g. for MC) looks like that:

process = PFnano_customizeMC(process)
#process = PFnano_customizeMC_add_DeepJet(process)                  ##### DeepJet inputs are added to the Jet collection
#process = PFnano_customizeMC_allPF(process)                        ##### PFcands will content ALL the PF Cands
#process = PFnano_customizeMC_allPF_add_DeepJet(process)            ##### PFcands will content ALL the PF Cands; + DeepJet inputs for Jets
#process = PFnano_customizeMC_AK4JetsOnly(process)                  ##### PFcands will content only the AK4 jets PF cands
#process = PFnano_customizeMC_AK4JetsOnly_add_DeepJet(process)      ##### PFcands will content only the AK4 jets PF cands; + DeepJet inputs for Jets
#process = PFnano_customizeMC_AK8JetsOnly(process)                  ##### PFcands will content only the AK8 jets PF cands
#process = PFnano_customizeMC_noInputs(process)                     ##### No PFcands but all the other content is available.

In general, whenever _add_DeepJet is specified (does not apply to AK8JetsOnly and noInputs), the DeepJet inputs are added to the Jet collection. For all other cases that involve adding tagger inputs, only DeepCSV and / or DDX are taken into account as default (= the old behaviour when keepInputs=True). Internally, this is handled by selecting a list of taggers, namely choosing from DeepCSV, DeepJet, and DDX (or an empty list for the noInputs-case, formerly done by setting keepInputs=False, now set keepInputs=[]). This refers to a change of the logic inside pfnano_cff.py and addBTV.py. If one wants to use this new flexibility, one can also define new customization functions with other combinations of taggers. Currently, there are all configurations to reproduce the ones that were available previously, and all configuations that extend the old ones by adding DeepJet inputs. DeepJet outputs, on top of the discriminators already present in NanoAOD, are added in any case where AK4Jets are added, i.e. there is no need to require the full set of inputs to get the individual output nodes / probabilities. The updated description using PFnano_customizeMC_add_DeepJet can be viewed here and the size here.

The latest addition before moving to the Run3 recipe was the inclusion of a fine-grained truth flavour branch, which encodes various bits known from DeepNTuples into one integer. It can be activated for simulated samples by adding the _and_Truth flag to the customizer, i.e. using PFnano_customizeMC_add_DeepJet_and_Truth.

How to create python files using cmsDriver

All python config files were produced with cmsDriver.py.

Two imporant parameters that one needs to verify in the central nanoAOD documentation are --conditions and --era.

Pre UL cmsRun python config files are generated by running make_configs_preUL.sh

bash make_configs_preUL.sh  # run to only produce configs
bash make_configs_preUL.sh  -e # run to actually execute configs on 1000 events

UL cmsRun python config files are generated by running make_configs_UL.sh

bash make_configs_UL.sh  # run to only produce configs
bash make_configs_UL.sh  -e # run to actually execute configs on 1000 events

Submission to CRAB

For crab submission a handler script crabby.py, a crab baseline template template_crab.py and an example submission yaml card card_example.yml are provided.

  • A single campaign (data/mc, year, config, output path) should be configured statically in a copy of card_example.yml.
  • To submit:
    source crab.sh
    python crabby.py -c card.yml --make --submit
    
  • --make and --submit calls are independent, allowing manual inspection of submit configs
  • Add --test to disable publication on otherwise publishable config and produce a single file per dataset
If experiencing problems with crab submission using the above instructions, e.g. on NAF-DESY
```
source /cvmfs/grid.cern.ch/centos7-umd4-ui-4_200423/etc/profile.d/setup-c7-ui-example.sh
source /cvmfs/cms.cern.ch/common/crab-setup.sh prod
source /cvmfs/cms.cern.ch/cmsset_default.sh
< navigate to CMSSW_X_Y_Z/src >
cmsenv
cd PhysicsTools/PFNano/test
```


and proceed with `crabby.py` as explained above, activate proxy for submission to be successful.
Useful commands to get paths to individual processed files
```
xrdfs [insert redirector to site] ls /store/path/to/your/crab/output/serialnumber > filelist.txt
( if there is more than one serial number (more than 1k files processed) repeat command but append to textfile using >> instead of > )
( clean textfile for log entries )
( then append the redirector (needs modification by you for specific site) using this helper )
python dataset_paths.py name_of_txt_file T2_DE_RWTH
```
Deprecated submission. Samples can be submitted to crab using the `submit_all.py` script. Run with `-h` option to see usage. Example can look like this:
```
python submit_all.py -c nano_config.py -s T2_DE_RWTH -f datasets/text_list.txt  -o /store/user/$USER/PFNano/  --ext test --test -d crab_noinpts

```

For the UL datasets:
```
##python submit_all.py -c nano102x_on_mini94x_2016_mc_NANO.py  -f 2016mc_miniAODv3_DY.txt  -d NANO2016MC

python submit_all.py -c nano106Xv8_on_mini106X_2017_mc_NANO.py -f 2017mc_miniAODv2_DY.txt  -d NANO2017MC

python submit_all.py -c nano106Xv8_on_mini106X_2018_mc_NANO.py -f 2018mc_DY.txt  -d NANO2018MC


##python submit_all.py -c nano102x_on_mini94x_2016_data_NANO.py -f 2016data_17Jul2018.txt -d NANO2016 -l Cert_271036-284044_13TeV_23Sep2016ReReco_Collisions16_JSON.txt

python submit_all.py -c nano106Xv8_on_mini106X_2017_data_NANO.py -f 2017data_31Mar2018.txt  -d NANO2017 -l /afs/cern.ch/cms/CAF/CMSCOMM/COMM_DQM/certification/Collisions17/13TeV/Legacy_2017/Cert_294927-306462_13TeV_UL2017_Collisions17_GoldenJSON.txt 

python submit_all.py -c nano106Xv8_on_mini106X_2018_data_NANO.py -f datasets_2018D.txt -d NANO2018 -l /afs/cern.ch/cms/CAF/CMSCOMM/COMM_DQM/certification/Collisions18/13TeV/Legacy_2018/Cert_314472-325175_13TeV_Legacy2018_Collisions18_JSON.txt 
```

Processing data

When processing data, a lumi mask should be applied. The so called golden JSON should be applicable in most cases. Should also be checked here https://twiki.cern.ch/twiki/bin/view/CMS/PdmV

  • Golden JSON, UL
# 2017: /afs/cern.ch/cms/CAF/CMSCOMM/COMM_DQM/certification/Collisions17/13TeV/Legacy_2017/Cert_294927-306462_13TeV_UL2017_Collisions17_GoldenJSON.txt
# 2018: /afs/cern.ch/cms/CAF/CMSCOMM/COMM_DQM/certification/Collisions18/13TeV/Legacy_2018/Cert_314472-325175_13TeV_Legacy2018_Collisions18_JSON.txt
#
  • Golden JSON, pre-UL
# 2016
jsons/Cert_271036-284044_13TeV_23Sep2016ReReco_Collisions16_JSON.txt
# 2017 
jsons/Cert_294927-306462_13TeV_EOY2017ReReco_Collisions17_JSON_v1.txt
# 2018
jsons/Cert_314472-325175_13TeV_17SeptEarlyReReco2018ABC_PromptEraD_Collisions18_JSON.txt

Include in card.yml for crabby.py submission. (In deprecated interactive submissiong add --lumiMask jsons/...txt)

How to create website with nanoAOD content

To create nice websites like this one with the content of nanoAOD, use the inspectNanoFile.py file from the PhysicsTools/nanoAOD package as:

python PhysicsTools/NanoAOD/test/inspectNanoFile.py NANOAOD.root -s website_with_collectionsize.html -d website_with_collectiondescription.html

Documenting the Extended NanoAOD Samples

Please document the input and output datasets on the following twiki: https://twiki.cern.ch/twiki/bin/view/CMS/JetMET/JMARNanoAODv1. For the MC, the number of events can be found by looking up the output dataset in DAS. For the data, you will need to run brilcalc to get the total luminosity of the dataset. See the instructions below.

Running brilcalc

These are condensed instructions from the lumi POG TWiki (https://twiki.cern.ch/twiki/bin/view/CMS/TWikiLUM). Also see the brilcalc quickstart guide: https://twiki.cern.ch/twiki/bin/viewauth/CMS/BrilcalcQuickStart.

Note: brilcalc should be run on lxplus. It does not work on the lpc.

Instructions:

1.) Add the following lines to your .bashrc file (or equivalent for your shell). Don't forget to source this file afterwards!

export PATH=$HOME/.local/bin:/cvmfs/cms-bril.cern.ch/brilconda/bin:$PATH
export PATH=/afs/cern.ch/cms/lumi/brilconda-1.1.7/bin:$HOME/.local/bin:$PATH

2.) Install brilws:

pip install --install-option="--prefix=$HOME/.local" brilws

3.) Get the json file for your output dataset. In the area in which you submitted your jobs:

crab report -d [your crab directory]

The processedLumis.json file will tell you which lumi sections you successfully ran over. The lumi sections for incomplete, failed, or unpublished jobs are listed in notFinishedLumis.json, failedLumis.json, and notPublishedLumis.json. More info can be found at https://twiki.cern.ch/twiki/bin/view/CMSPublic/CRAB3Commands#crab_report.

4.) Run brilcalc on lxplus:

brilcalc lumi -i processedLumis.json -u /fb --normtag /cvmfs/cms-bril.cern.ch/cms-lumi-pog/Normtags/normtag_PHYSICS.json -b "STABLE BEAMS"

The luminosity of interest will be listed under "totrecorded(/fb)." You can also run this over the other previously mentioned json files.

Note: '-b "STABLE BEAMS"' is optional if you've already run over the golden json. Using the normtag is NOT OPTIONAL, as it defines the final calibrations and detectors that are used for a given run.