README.md

metadata

title: MolCraftDiffusion - 3D Molecule Generator
emoji: 🧪
colorFrom: purple
colorTo: blue
sdk: gradio
sdk_version: 4.44.0
python_version: '3.10'
app_file: app.py
pinned: false
license: mit

MolCraftDiffusion Demo

Generate novel 3D molecular structures using a pre-trained Equivariant Diffusion Model (EDM).

This demo uses the MolecularDiffusion framework.

How to use

1. Select the number of molecules you want to generate.
2. Choose the size (number of atoms) or let the model sample from the training distribution.
3. Adjust diffusion steps (more steps = usually better quality but slower).
4. Click Generate.

The molecules will be visualized interactively in the browser using py3Dmol. You can also download the generated .xyz files.

Model Details

• Architecture: Equivariant Graph Diffusion Model (EGCL)
• Training Data: QM9 / GEOM (Drug-like molecules)
• Input: Gaussian noise
• Output: 3D coordinates + atom types

Credits

Powered by MolecularDiffusion.