AGeDi Documentation

AGeDi is an atomistic generative diffusion package for training and sampling structure generators from ASE datasets.

Documentation

• Overview
  • What AGeDi is
  • Core capabilities
  • High-level package layout
  • Typical workflow
• Installation
  • Requirements
  • Install from GitHub
  • Developer install
  • Verify installation
• Concepts and model behavior
  • Graph representation
  • Diffusion components
  • Position noisers
  • Sampling semantics
  • Training outputs
  • Property conditioning
  • Data augmentation (cell repeat)
• Usage guide
  • Command-line interface
  • Python API workflow
  • YAML-based training
  • Object-based workflow
  • PdO-on-Pd end-to-end example
• Pitfalls and troubleshooting
  • Sampling input pitfalls
  • Training pitfalls
  • Data pitfalls
  • Operational checks
• References and citation
  • Primary publications
  • How this documentation maps to the papers
  • Suggested citation
• API Reference
  • agedi