deepmodeling/deepmd-kit

Supports multi-backend training (TensorFlow, PyTorch, JAX, Paddle) and integrates with major MD engines (LAMMPS, GROMACS, OpenMM, AMBER, CP2K, i-PI, ABACUS) via unified model export. Implements the Deep Potential architecture family, which encodes system symmetries through local atomic environments and sub-networks to compute additive atomic energies, enabling accurate interatomic potentials that scale linearly with system size while remaining orders of magnitude faster than ab initio methods.

1,892 stars and 7,196 monthly downloads. Used by 2 other packages. Actively maintained with 52 commits in the last 30 days. Available on PyPI and npm.