deepmodeling/deepmd-kit

A deep learning package for many-body potential energy representation and molecular dynamics

What it solves

DeePMD-kit converts quantum-mechanical reference data into fast, scalable interatomic potentials. This allows researchers in molecular and materials science to perform large-scale simulations of finite molecules, covalent systems, periodic solids, and metals without the computational cost of full quantum mechanical calculations.

How it works

The toolkit provides a workflow to either train a model from scratch or start with a pretrained Deep Potential (DPA4) model and fine-tune it for a specific system. It supports multiple backends including TensorFlow, PyTorch, JAX, and Paddle. Once trained, models can be compressed for faster inference and exported to be used in various molecular dynamics engines.

Who it’s for

It is designed for scientists and engineers in molecular and materials science who need high-performance interatomic potentials for molecular dynamics simulations.

Highlights

  • Pretrained-first workflows: Support for downloading and fine-tuning DPA4 models.
  • Flexible Model Portfolio: Offers DPA4 for high accuracy and DPA4C for high throughput.
  • Broad Physical Targets: Models energy, forces, virials, Hessians, spin, magnetic forces, dipoles, and electronic density of states.
  • Backend Agnostic: Compatible with PyTorch, TensorFlow, JAX, and Paddle.
  • Extensive Integration: Connects with simulation engines like LAMMPS, i-PI, ASE, GROMACS, and CP2K.

Related

  • Project
    atomistic-machine-learning/schnetpack

    A deep learning toolbox for predicting quantum-chemical properties of molecules and materials using atomistic neural networks.

  • Project
    microsoft/mattersim

    A deep learning atomistic model for simulating bulk materials across various elements, temperatures, and pressures to predict energy, forces, and stress.

  • Project
    deepqmc/deepqmc

    An open-source software suite for the variational optimization of deep-learning molecular wave functions to simulate electronic ground and excited states.

  • Project
    deepmodeling/dpgen

    A concurrent learning platform that automates the generation of deep learning-based potential energy models and force fields for molecular simulations on HPC clusters.

  • Project
    brucefan1983/GPUMD

    A highly efficient, GPU-accelerated molecular dynamics package that enables the training and using of neuroevolution machine-learned potentials for large-scale atomistic simulations.