pyrfp 0.1.2

Python package for the data-driven Rosenbluth-Fokker-Planck equation.

Python Package for the Rosenbluth Fokker Planck (RFP) Equation

Currently, heavily under renovation (refactoring) from my old code (in my other private repository)

Description

This package is refactored version of a part of the pystops_ml code. I've separated only data generation and training part from pystops_ml.

Unlike pystops_ml, this module doesn't utilize distributed training. (DDP feature not needed)

This code is part of my paper, Data-Driven Stochastic Particle Scheme for Collisional Plasma Simulations.

• Preprint is available at SSRN

Features

• Data generation
  • Uncertainty quantification using the maximum entropy distribution (using pymaxed)
  • Axisymmetric evaluation of the Rosenbluth potentials and their derivatives
• Data training: supports cpu, cuda, and mps training
• Data-driven simulation: rfp-ann

Installation

You can install via pip

$python3 -m pip install pyrfp

Dependencies

• Global

• python >=3.10

• torch >= 1.13.1

• Personal project

  • pymaxed (for the Maximum Entropy Distribution)
  • pyapes (for the field solver)
  • pymytools (miscellaneous tools including data IO, logging, and progress bar)