ai²-kit

A toolkit featured artificial intelligence × ab initio for computational chemistry research.

Please be advised that ai2-kit is still under heavy development and you should expect things to change often. We encourage people to play and explore with ai2-kit, and stay tuned with us for more features to come.

Feature Highlights

• Collection of tools to facilitate the development of automated workflows for computational chemistry research.
• Utilities to execute and manage jobs in local or remote HPC job scheduler.
• Utilities to simplified automated workflows development with reusable components.

Installation

You can use the following command to install ai2-kit:

# for users who just use most common features
pip install ai2-kit

# for users who want to use all features
pip install ai2-kit[all]

If you want to run ai2-kit from source, you can run the following commands in the project folder:

pip install poetry
# If you meet ConnectionError:
# poetry config installer.max-workers 4
poetry install

./ai2-kit --help

Note that instead of running global ai2-kit command, you should run ./ai2-kit to run the command from source on Linux/MacOS or .\ai2-kit.ps1 on Windows.

Manuals

Featuring Tools

• ai2-cat: A toolkit for dynamic catalysis researching.

Workflows

• CLL MLP Training Workflow (zh)

Domain Specific Tools

• Proton Transfer Analysis Toolkit
• Amorphous Oxides Structure Analysis Toolkit

General Tools

• Tips: useful tips for using ai2-kit
• Batch Toolkit: a toolkit to generate batch scripts from files or directories
• ASE Toolkit: commands to process trajectory files with ASE
• DPData Toolkit: commands to process system data with dpdata
• Model Deviation Toolkit: a toolkit to filter structures by model deviation

Notebooks

• ai2cat

Miscellaneous

• HPC Executor Introduction (zh): a simple HPC executor for job submission and management
• Checkpoint Mechanism

Acknowledgement

This project is inspired by and built upon the following projects:

• dpgen: A concurrent learning platform for the generation of reliable deep learning based potential energy models.
• ase: Atomic Simulation Environment.