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. For more information, please refer to the documentation.

• Hybrid and Modular Neural Network and Classical Potential;
• High-Performance C++ Operators;
• Built-in torch-based MD simulation;
• Versatile Data Pipeline;
• User-Friendly App Wrapping

To install molpot, you can use pip:

pip install molpot

or clone the repository and install it manually:

git clone http:
cd molpot
pip install .

or install from docker

docker pull molpot/molpot:latest

or apptainer by using definition file

apptainer build molpot.sif molpot.def

Instead of giving getting started guide here, we prefer to give a tutorial about how to learn this package, since molpot is a general and extensible package. Once you have a requirements, find a similar example or tutorial, read the code and check the docstring, and you will get a good understanding. Here is a table of content of the doc:

• tutorial: A how-to notebook driven by tasks
• examples: A collection of examples
• getting started: Introduce the basic concepts
• API: A complete API reference
• contributing: how to customize your own module

TO BE ANNOUNCED

• Physical-based potential support
• Active learning