Introduction

Transform Your World (TYW) is a project for creating and using transformer models for materials discovery.

Installation

Run

bash full_install.sh

which will install Pixi, a Pixi build of TYW, Pixi builds for sub-packages such as alfabet, and sets environment variables based on the install location.

Manual Installation

First install pixi onto your machine (see the wiki for this project or the Pixi documentation for how to do this).

Next, while in a directory for TYW, run

pixi install 

to install default required dependencies and

pixi shell 

to install the dependencies and to activate the environment. For GPU acceleration, run

pixi install --environment gpu 

This may then be run via

pixi run --frozen --environment gpu python <example_script>.py

If running across multiple nodes on a SLURM cluster, modify the above to:

pixi run --frozen --environment gpu srun python <example_script>.py

You will need to install subpackages such as those in alfabet yourself afterwards as well as set environment variables as desired.

Usage

transform_your_world is designed to be used as a machine learning library for use in (relatively) lightweight Python scripts. The user guide for this project also outlines typical use patterns.

Tests

Tests are configured to be run via pytest. Run tests (with code coverage check) via

pytest --cov 

User Guide and API Documentation

See the HTML files in /docs. Docs may be generated via

pixi shell 
sphinx-build -M html docs docs/_build

The full documentation contains a high-level user guide as well as comprehensive API documentation. It is recommended to start with the former.

For training models, see transform_your_world.experiment.

For data preprocessing and setting up data pipelines, see transform_your_world.features and transform_your_world.datasets to get started.

For developing models, transform_your_world.model and transform_your_world.training are most relevant.

For generating new molecules and assessing model performance, see transform_your_world.generation and transform_your_world.benchmarking.

Models

A suite of chemistry GPT models built using Transform Your World are available at , from which one example is mirrored on Hugging Face.

Copyright

This project is MIT-licensed (see LICENSE.txt). It has been approved for open source release under O4996.

© 2026. Triad National Security, LLC. All rights reserved.

This program was produced under U.S. Government contract 89233218CNA000001 for Los Alamos

National Laboratory (LANL), which is operated by Triad National Security, LLC for the U.S.

Department of Energy/National Nuclear Security Administration. All rights in the program are

reserved by Triad National Security, LLC, and the U.S. Department of Energy/National Nuclear

Security Administration. The Government is granted for itself and others acting on its behalf a

nonexclusive, paid-up, irrevocable worldwide license in this material to reproduce, prepare

derivative works, distribute copies to the public, perform publicly and display publicly, and to permit

others to do so.