Warning

Not ready yet. Use with caution.

A High Granularity Toy Calorimeter

The source code for a toy calorimeter to study particle reconstruction performance.

This calorimeter was also employed in the following paper: https://link.springer.com/article/10.1140/epjc/s10052-022-10665-7

Cite as:

@article{Qasim_2022,
	doi = {10.1140/epjc/s10052-022-10665-7},
	url = {https://doi.org/10.1140%2Fepjc%2Fs10052-022-10665-7},
	year = 2022,
	month = {aug},
	publisher = {Springer Science and Business Media {LLC}},
	volume = {82},
	number = {8},
	author = {Shah Rukh Qasim and Nadezda Chernyavskaya and Jan Kieseler and Kenneth Long and Oleksandr Viazlo and Maurizio Pierini and Raheel Nawaz},
	title = {End-to-end multi-particle reconstruction in high occupancy imaging calorimeters with graph neural networks},
	journal = {The European Physical Journal C}
}

Install / Compile

The following docker container should be used to run the code in this repository: https://hub.docker.com/r/shahrukhqasim2/g4calo. The repository has to be manually compiled within the container. A container with a prebuilt package is coming soon.

The singularity .def and Dockerfile are both available in containers/ directory if the container needs to be built.

singularity pull docker://shahrukhqasim2/g4calo:latest

This step can take many hours to complete. As the result, a .sif file will be created in the current directory. Login to the .sif container:

singularity shell g4calo_latest.sif

If at CERN, EOS and AFS can also be mounted as follows:

singularity shell -B /afs -B /eos g4calo_latest.sif

Check out the repository:

git clone https://github.com/shahrukhqasim/hg_toy_calorimeter.git
cd hg_toy_calorimeter
git submodule update --init --recursive

And build:

mkdir build
cd build
cmake ..
make -j
cd ..

Generating data

As described in the paper, there are two steps to generate a datset with this calorimeter. First individual simulations are generated which can be either single particle simulations or proton-proton interactions. A dataset of simulations is generated and then to generate events, simulations are sampled for every event without replacement.

Setting environment:

To set the environment once the code has been built, the build directory should be added to PYTHONPATH.

cd python/bin
export PYTHONPATH=$PYTHONPATH:`readlink -f ../../build/`:`readlink -f ../../build/lib/`:`readlink -f ../libs`
cd ../..

Generate simulations

run_simulation.py script can be used to generate a set of simulations:

cd python/bin
python3 run_simulations.py minbias --cores=1

The simulations will be generated in ra_pickles format, as described here: https://github.com/shahrukhqasim/ra_pickles

They are inherently stored as pickled numpy arrays and therefore, can be accessed outside the container, by only installing the ra_pickles package via pip3 install ra-pickles. The container should only be used for gerating events and simulations.

Generate events

generate_events.py script can be used to generate a set of events:

cd python/bin
python3 create_dataset.py example_datasets_config.ini test_dataset_cern

Here, first the HGCAL environment must have been sourced, to be able to export data in .djctd format.

Events export in ra_pickles format is coming soon.