Caloscore v2 official repository

Official repository for CaloScore v2, an update to CaloScore that uses a diffusion generative model for fast detector simulation with single-shot sampling!

Requirements

Packages used for training and sampling are found in the requirements.txt text file and can be directly installed with pip.

Data

Results are presented using the Fast Calorimeter Data Challenge dataset and are available for download on zenodo:

Run the training scripts with

cd scripts
python train.py  --config CONFIG

CONFIG options are [config_dataset1.json/config_dataset2.json/config_dataset3.json]

After training the baseline model you can run the progressive distillation with the commands:

cd scripts
python train.py  --config CONFIG --distill --factor 2

For additional distillation steps just multiply --factor by a power of 2.

Sampling from the learned score function

python plot_caloscore.py  --sample  --config CONFIG [--distill] [--factor 2]

Where again factor and distill flags are used to load the distilled models instead.

Creating the plots shown in the paper

python plot_caloscore.py  --config CONFIG [--distill] [--factor 2]