caloclouds-2

PyTorch implementation of the CaloClouds II model introduced in https://arxiv.org/abs/2309.05704

https://github.com/flc-qu-hep/caloclouds-2

Science Score: 54.0%

This score indicates how likely this project is to be science-related based on various indicators:

  • CITATION.cff file
    Found CITATION.cff file
  • codemeta.json file
    Found codemeta.json file
  • .zenodo.json file
    Found .zenodo.json file
  • DOI references
  • Academic publication links
    Links to: arxiv.org
  • Academic email domains
  • Institutional organization owner
  • JOSS paper metadata
  • Scientific vocabulary similarity
    Low similarity (7.2%) to scientific vocabulary
Last synced: 10 months ago · JSON representation ·

Repository

PyTorch implementation of the CaloClouds II model introduced in https://arxiv.org/abs/2309.05704

Basic Info
  • Host: GitHub
  • Owner: FLC-QU-hep
  • License: apache-2.0
  • Language: Jupyter Notebook
  • Default Branch: main
  • Homepage:
  • Size: 1.39 MB
Statistics
  • Stars: 3
  • Watchers: 1
  • Forks: 3
  • Open Issues: 0
  • Releases: 0
Created almost 3 years ago · Last pushed over 2 years ago
Metadata Files
Readme License Citation

README.md

CaloClouds II Model for Fast Calorimeter Simulation

arXiv License

PyTorch implementation of the CaloClouds II Model introduced in CaloClouds II: Ultra-Fast Geometry-Independent Highly-Granular Calorimeter Simulation (arXiv:2309.05704).

The model is an evolution of the CaloClouds model published in (arXiv:2305.04847).


The CaloClouds II Model generates photon showers in point cloud format with up to 6000 points per shower for an energy range between 10 and 90 GeV. It consists out of multiple sub-generative models, including the PointWise Net trained as a score-based generative model and a normalizing flow for generating shower observables. Further, consistency distllation is implemented to distill a trained CaloClouds II diffusion model into a consistency model dubbed CaloClouds II (CM). The training data is generated with a Geant4 simulation of the planned electromagnetic calorimeter of the International Large Detector (ILD).


The CaloClouds II Pointwise Net Diffusion Model can be trained via python training.py using the default parameters set in config.py.

A trained CaloClouds II Diffusion model can be distilled into a consistency model via python distillation.py.

The Shower Flow is trained via the notebook ShowerFlow.ipynb.

The polynomial fits for the occupancy calculations are performed in occupancy_scale.ipynb.

An outline of the sampling process for both CaloClouds II and CaloClouds II (CM) can be found in generate.py.

The timing of the models is benchmarked with calc_timing.py


The training dataset is available under the link: https://syncandshare.desy.de/index.php/s/XfDwx33ryERwPdi


Code references: - The code for training the score-based model is based on: https://github.com/crowsonkb/k-diffusion - The consistency distillation is based on: https://github.com/openai/consistency_models/ - The PointWise Net is adapted from: https://github.com/luost26/diffusion-point-cloud - Code base for our CaloClouds (1) model: https://github.com/FLC-QU-hep/CaloClouds

Owner

  • Name: FLC-QU-hep
  • Login: FLC-QU-hep
  • Kind: organization

Citation (CITATION.cff)

cff-version: 1.2.0
message: "The following software is a product of a published paper, if you use the software, please cite out paper as below."
authors:
    - family-names: Buhmann
      given-names: Erik
      orcid: https://orcid.org/0000-0002-4805-3721
    - family-names: Gaede
      given-names: Frank
      orcid: https://orcid.org/0000-0002-7055-9200
    - family-names: Kasieczka
      given-names: Gregor
      orcid: https://orcid.org/0000-0003-3457-2755
    - family-names: Korol
      given-names: Anatolii
      orcid: https://orcid.org/0000-0002-2569-1771
    - family-names: Korcari
      given-names: William
      orcid: https://orcid.org/0000-0001-8017-5502
    - family-names: Krüger
      given-names: Katja
      orcid: https://orcid.org/0000-0002-1956-6608
    - family-names: McKeown
      given-names: Peter
      orcid: https://orcid.org/0009-0006-9722-2233
title: "CaloClouds II: Ultra-Fast Geometry-Independent Highly-Granular Calorimeter Simulation"
version: 1.0
identifiers:
    - type: doi
    - value: 10.48550/arXiv.2309.05704
date-released: 2023-12-12
      

GitHub Events

Total
Last Year