deepmodeling/dpgen

The deep potential generator to generate a deep-learning based model of interatomic potential energy and force field

What it solves

DP-GEN automates the creation of reliable deep learning-based potential energy models and force fields for molecular simulations. It eliminates the need for manual effort in preparing scripts and managing job queues on High Performance Computing (HPC) clusters when generating these models.

How it works

DP-GEN uses a concurrent learning platform that integrates with DeePMD-kit to generate models. It samples millions of structures but selects only a small fraction for expensive first-principles calculations to ensure the final model is uniformly accurate. The software acts as an orchestrator that interfaces with various molecular simulation tools (like LAMMPS, Gromacs, and AMBER) and ab-initio calculation software (like VASP, CP2K, and Gaussian) to handle data generation and analysis.

Who it’s for

It is designed for researchers and developers in computational chemistry and physics who need to generate accurate interatomic potential energy models using deep learning on HPC systems.

Highlights

  • Automated HPC Management: Automatically prepares scripts and maintains job queues for Slurm, PBS, LSF, and cloud machines.
  • High Efficiency: Capable of sampling tens of millions of structures while minimizing the number of required first-principles calculations.
  • Broad Compatibility: Supports a wide array of MD interfaces (LAMMPS, Gromacs, AMBER, Calypso) and ab-initio tools (VASP, PWSCF, CP2K, SIESTA, Gaussian, Abacus, PWmat).
  • Modular Design: Features a modular code structure that allows users to extend the platform for specific needs.

Related

  • Project
    deepmodeling/deepmd-kit

    A deep learning package that turns quantum-mechanical reference data into fast, scalable interatomic potentials for molecular and materials science simulations.

  • Project
    hpcaitech/ColossalAI

    A distributed deep learning framework that makes training and inference for large AI models faster and cheaper through advanced parallelism and memory management.

  • Project
    uxlfoundation/oneDAL

    A C++ and DPC++ library that provides hardware-accelerated machine learning routines for tabular data across CPUs, GPUs, and distributed systems.

  • Project
    materialyzeai/matgl

    A graph deep learning library for materials science that provides various GNN architectures to predict material properties and simulate potential energy surfaces.

  • Project
    mosaicml/composer

    A PyTorch-based deep learning library that simplifies distributed training at scale, providing tools for memory optimization and algorithmic speedups to reduce training time and cost.