Skip to content

LonxunQuantum/lammps-MatPL

Repository files navigation

MatPL for lammps

├── dp_lmps_demo/
├── LICENSE
├── lmp_for_cmake/
├── lmp_nepkokkos_cmake/
├── Makefile.mpi
├── MATPL/
├── nep_lmps_demo/
└── README.md
  • lmp_nepkokkos_cmake is the LAMMPS force field interface for MatPL-2026.3, supporting NEP KOKKOS GPU acceleration, NEP CPU version, and DP (CPU and GPU acceleration). It is compiled using CMake.

  • MATPL and Makefile.mpi are the LAMMPS force field interface for MatPL-2025.3, supporting NEP (CPU and GPU acceleration) and DP (CPU and GPU acceleration). They are compiled using Make. nep_lmps_demo and dp_lmps_demo are MD test examples for the NEP and DP force fields of MatPL-2025.3, respectively.

  • The key advancement in the 2026.3 version is its use of KOKKOS to offload neighbor list construction to the GPU. This fundamental change, combined with further kernel optimizations, accelerates the NEP GPU acceleration by over an order of magnitude compared to the 2025.3 version.

  • The MatPL-2026.3 LAMMPS interface requires the LAMMPS source code version not to exceed 2024.

For installation instructions, please refer to http://doc.lonxun.com/MatPL/install/

Citation

https://chemrxiv.org/doi/full/10.26434/chemrxiv.15001665

About

The LAMMPS interface for the machine learning force field of MatPL.

Resources

License

Stars

Watchers

Forks

Packages

 
 
 

Contributors