Package to optimize MT2 Analysis with ML-based methods
Will contains several tools:
- to convert ROOT ntuples into Numpy Array
- to train model on data (Keras) and evaluate model
- to write back ROOT ntuples with trained model is evaluated
Log in to t3ui02 -> this enables access to MT2 ntuples in /scratch As soon as you want to submit a job to GPU via slurm, you will need to login into t3ui04 and find a solution to get ntuples accessible
Environment (uproot, keras, numpy)
export PATH=/t3home/mratti/miniconda3/bin:$PATH
conda activate tensorflow_base
OUTDATED: Environment (enables uproot):
source /work/mratti/bootAnaconda_fromMauro.sh
To activate tensorflow environment on CPU or GPU:
conda env list
source activate tensorflow
source activate tensorflow_gpu
END OUTDATED
To run a jupyter notebook:
jupyter notebook --port 8883 --no-browser
To display the notebook on your lapton browser
ssh -N -f -L localhost:8883:localhost:8883 t3ui02.psi.ch
http://localhost:8883/tree
Installation:
git clone git@github.com:MT2Analysis/MT2MLO.git
git checkout -b <own-branch>
Development done in own branch, then PR to master for review and merging:
git add bla.py
git commit -m "reasonable comment"
git push origin <own-branch>
cd rootToNumpy
python convertRootToNumpy.py
cd ../models
python example_V01.py
cd rootToNumpy
python evaluateAndAttach.py