fitsnap Goto Github PK

A Python package for machine learning potentials with LAMMPS.

Documentation page: https://fitsnap.github.io

Colab Python notebook tutorial: https://colab.research.google.com/github/FitSNAP/FitSNAP/blob/master/tutorial.ipynb

Rohskopf et al., (2023). FitSNAP: Atomistic machine learning with LAMMPS. Journal of Open Source Software, 8(84), 5118, https://doi.org/10.21105/joss.05118

• This package expects Python 3.10+
• Python dependencies: See pyproject.toml
• Compile LAMMPS as a shared library with python support. If you can run import lammps; lmp = lammps.lammps() without errors in your Python interpreter, you're good to go!
• [Optional] To use neural network fitting functionality, install PyTorch.
• [Optional] For optimal performance, also install your favorite flavor of MPI (OpenMPI, MPICH) and the Python package mpi4py. If installing mpi4py with a Python package manager, we recommend using pip over conda as pip will auto-configure your package to your system's defaut MPI version (usually what you used to build LAMMPS).

WARNING: Conda LAMMPS installation does NOT include ACE. See the docs for details on how to install the current LAMMPS which has these functionalities.

• Add conda-forge to your conda install, if not already added:
  conda config --add channels conda-forge
• Create a new conda environment:
  conda create -n fitsnap python=3.9; conda activate fitsnap;
• Install the following packages:
  conda install -c conda-forge lammps fitsnap3

• (mpirun -np #) python -m fitsnap3 [options] infile
• Command line options can be seen with python -m fitsnap3 -h
• Examples of published SNAP interatomic potentials are found in examples/
• Examples of running FitSNAP via the library interface are found in examples/library

• See our Programmer Guide on how to add new features.
• Abide by our code standards by installing flake8 and running flake8 --statistics in the top directory.
• Get Sphinx with pip install sphinx sphinx_rtd_theme for adding new documentation, and see docs/README.md for how to build docs for your features.
• Feel free to ask for help!

• Mitchell Wood and Aidan Thompson co-lead development of FitSNAP since 2016.
• The FitSNAP Development Team is the set of all contributors to the FitSNAP project, including all subprojects.
• The core development of FitSNAP is performed at the Center for Computing Research (CCR), Sandia National Laboratories, Albuquerque, New Mexico, USA
• The original prototype of FitSNAP was developed in 2012 under a CIS LDRD project.

Copyright (2016) Sandia Corporation. Under the terms of Contract DE-AC04-94AL85000 with Sandia Corporation, the U.S. Government retains certain rights in this software. This software is distributed under the GNU General Public License