Skip to content

Molpot User Manual

molpot is designed for efficient building, training and deployment of molecular simulation potentials, leveraging a modular design to enable flexible customization and seamless integration into different workflows. It incorporates cutting-edge techniques, such as torch.compile for optimized deep learning execution and high-performance C++ operators for computational efficiency.

Key Features

  • Modular Design: Easily combine different types of potential functions.
  • Efficient Training: Supports parallel computing and GPU acceleration.
  • Extensibility: Easily integrates new neural network models or physical potential functions.
  • Physics-Guided Learning: Incorporates physical constraints to improve interpretability and generalization.
  • Compatibility: Supports multiple data formats and interfaces with existing molecular simulation software.

Reading Guide