Parnassus: A GPU-enabled, Python-based Package for Fast Particle Detector Simulation and Reconstruction
📄 arXiv:2606.26089 · 📥 PDF · 2026-06-24 · hep-ph
Authors: Abdelrahman Elabd [arXiv · scholar] , Eilam Gross [arXiv · scholar] , Dmitrii Kobylianskii [arXiv · scholar] , Benjamin Nachman [arXiv · scholar]
🕰 Orloj analysis
Parnassus je nový, veřejně dostupný software v Pythonu/PyTorch s podporou GPU pro rychlou simulaci a rekonstrukci detektorů v částicové a jaderné fyzice. Nabízí uživatelsky přívětivé rozhraní s vyměnitelnými modely detektorů, včetně neuronových sítí pro emulaci Geant4 a parametrických modelů ve stylu Delphes, bez nutnosti instalace ROOTu.
💡 Tento software představuje významný praktický pokrok v simulaci detektorů díky využití GPU, PyTorch a neuronových sítí, což zvyšuje efektivitu a dostupnost pro komunitu částicové fyziky.
Categories:
EXP-1
PRT-3
MET-3
MET-1
✓ code_available, falsifiable, modest_claims
📄 Abstract
We present the public software release of Parnassus, a Python/PyTorch, GPU-compatible framework for fast detector simulation and reconstruction in particle and nuclear physics. Parnassus provides a user-friendly framework with interchangeable detector models: neural models can emulate computationally expensive Geant4-based detector simulation and reconstruction chains, while parametric models provide PyTorch implementations of selected Delphes-style detector responses. This initial release includes two models of the CMS detector: one based on a flow-matching neural network architecture and one based on a PyTorch implementation of the Delphes CMS card (parametric bias and smearing). PyTorch versions of the ATLAS and ALEPH Delphes cards are also available, together with a flow-matching neural model of the ALEPH detector that extends the framework to the e+e- LEP environment. All detector-specific backends share the same process-agnostic and detector-agnostic API: users select a detector card - analogous to choosing a detector card in Delphes - and the same tool can be applied to new physics processes without retraining the released detector model. There are native interfaces to the event generator Pythia and the event clustering package FastJet. Unlike previous C++/ROOT-based tools, Parnassus provides GPU-capable PyTorch detector-response backends and requires no ROOT installation. We describe the installation, command-line and Python API, configuration system, and demonstrate the framework on Standard Model and BSM processes.