LAMMPS + DeePMD — молекулярная динамика с ML-потенциалами

★ 7.5 · general

lammps-deepmd is a Claude Code skill that runs LAMMPS molecular dynamics simulations using DeePMD-kit machine learning potentials, achieving near-DFT accuracy at scales from 10K to 1M atoms. It provides ready-to-use NVT and NPT input script templates, supports multi-model deviation with up to four models writing per-frame force deviations to model_devi.out for active learning workflows, and handles structure conversion via ASE or dpdata. A LAMMPS build with the DEEPMD package and a frozen .pb model file are required prerequisites. The skill is designed for researchers studying diffusion, phase transitions, or surface reactions who need large-scale ML-driven MD without relying on classical force fields.