Package to produce CaloL1 calibration of the trigger towers.
cmsrel CMSSW_13_0_0_pre2
cd CMSSW_13_0_0_pre2/src
cmsenv
git cms-init
git remote add cms-l1t-offline [email protected]:cms-l1t-offline/cmssw.git
git fetch cms-l1t-offline l1t-integration-CMSSW_13_0_0_pre2
git cms-merge-topic -u cms-l1t-offline:l1t-integration-v142
git clone https://github.com/cms-l1t-offline/L1Trigger-L1TCalorimeter.git L1Trigger/L1TCalorimeter/data
git clone [email protected]:jonamotta/CaloL1CalibrationProducer.git
git cms-checkdeps -A -a
scram b -j 12To produce the L1NTuples on Tier3, go in L1NtupleLauncher and run:
python submitOnTier3.py <options>Examples of launching commands can be found in submitOnTier3.sh.
After the production of the L1NTuples the production of the input files to the NNs is done by going to L1NtupleReader and running:
python3 batchMaker.py <options>this will batch the L1NTuples in .hdf5 files containing no more then N events each (N to be specified).
After the batching, crate the taglist file and put it inside the folder L1NtupleReader/inputBatches
After this the Padding of the chunky donut needs to be performed with:
python batchSubmitOnTier3.py <options>Examples of launching commands can be found in batchSubmitOnTier3.sh.
After this, need to merge the batches into one single file containing the input to the NNs, this is done with:
python3 batchMerger.py <options>this will create the following four output files that are to be used for the training of the NNs:
X_train.npzX_test.npzY_train.npzY_test.npz
When the four inputs files above are produced the model can be trained with:
python3 NNModelTraining<NNversions>.py <options>To produce the Scale Factors matrix, run:
python3 CalibrationFactor<tag>.py <options>To make the plots of the output of the NN, run:
python3 ModelPlots<tag>.py <options>