This repository is the collection of code for the ${\tau\tau}$ inariant mass inference to be used in the HH ${\to}$ bb ${\tau\tau}$ CMS analysis.
The code is a work in progress.

Installation

To install the correct environment use conda

conda env create -f mTT_env.yaml
conda activate mTT_env.yaml
conda env config vars set \
    LD_LIBRARY_PATH=$(realpath $(dirname $CONDA_EXE)/../lib):$CONDA_PREFIX/lib:$LD_LIBRARY_PATH \
    MTTFIT_HOME=$PWD
conda deactivate
conda activate mtt-fit

Alternatively, the mTT_env-snapshot.yaml dictionary contains all the required packages and version explicitly mentioned.

Steps

Create .h5 files

.h5 files are the standard input to the NN taining script. The create_h5.py script creates .h5 files containing the needed columns from the given input. The script assigns the is_train, is_test and is_valid labels to be used during training. The precision of the data is reduced to standard of half precision to avoid memory issues.
To create the .h5 files, run

python create_h5.py --input input_patter --output output_file --target target_var --tree tree_name --train-size T --valid-size V --min m --max M --threads t --features /path/to/cfg.py

where --input is the input file path (or glob pattern), --output is the output .h5 file name, --target is the target variables name, --tree is the input ROOT tree name, --train-size is the train sample size, --valid-size is the validation sample size, --min is the lower threshold for the target, --max is the higher threshold for the target, --threads is the number of threads to use, --featuresis the cfg file containing the FEATURES dictionary, The test sample size is deduced from the train and test samples size.

Notes

Forced dependencies

• ROOT v6.26.02 or greater is needed
• Numpy 1.21.4 or lower is needed by Numba

Changes in the conda environment

When updating the conda enviroment file update also the snapshot dictionaty and be sure not to put any user specific path inside the dictionary. Run something like

conda activate mtt-fit
conda env export | sed "s_prefix:.*__g" | sed "s_variables:.*__g" | sed "s_LD\_LIBRARY\_PATH.*__g" | sed "s_MTTFIT\_HOME.*__g" > mTT_env-snapshot.yaml

to remove lines containing user paths.