Graphormer

Graphormer

Trains custom deep learning models for molecule tasks such as finding new materials or drugs. Aimed at researchers and developers, with checkpoint folders from two published studies.

Made by microsoft on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face.

About

What it is: Graphormer is a deep learning package for training your own models on molecule tasks. It aims to speed up AI research for molecule science, for example finding new materials or drugs. This repository holds notes on the project and two sets of model checkpoints.

Who it suits: It suits researchers and developers who want to train models on molecule data. The card is short, so expect to read the project website for the details.

Industry | Who uses it | How they use it | Example
Pharmaceuticals | Computational chemists | Train custom models on their own molecule data | Fit a model to a compound screening question
Materials science | Materials researchers | Apply the package to molecule modeling for new materials | Compare candidate materials with a trained model
Energy storage | Battery researchers | Start from the redox potential checkpoints | Reuse the orfb-design checkpoints in a flow battery study
Academia | Machine learning researchers | Study checkpoints from the published equilibrium distribution work | Open the Distributional-Graphormer folder and inspect the files

Anyone who needs a finished app with no Python work can skip it.

Where to use it: You work with it in a Python setup on your own machine.

• Windows, macOS or Linux with Python, to fetch the files
• Hugging Face, where the checkpoints are hosted
• Azure Quantum Elements, for the advanced pre-trained versions

Use case | Who | What they get back
Train a custom molecule model | Researchers and developers | A model fitted to their own molecule data
Study published results | Academic researchers | Checkpoints for the equilibrium distribution paper
Predict redox potential | Battery researchers | Checkpoints from the flow battery screening study

How to set it up: Setup is a file download, nothing more.

• Python on Windows, macOS or Linux
• Enough disk space for the checkpoint folders

1. Install the Hugging Face hub tools.
2. Download the repository files to your computer.
3. Open the checkpoint folder you want.

pip install -U huggingface_hub
hf download microsoft/Graphormer

How to use it: The card shows no run command, so day-to-day use starts from the files you downloaded. Two folders hold checkpoints, and the project website is the place to look for how the package itself is run.

• Distributional-Graphormer: checkpoints for the paper on equilibrium distributions of molecular systems
• orfb-design: checkpoints for redox potential prediction

Example: You want the redox potential checkpoints from the flow battery work. You run the download command, then open the folder named orfb-design among the files it fetches. The card gives no code to load them, so you follow the project website from there.

For newcomers and experts: Think of it as a research toolkit for teaching a computer to model molecules. A small first step is to download the files and look through the two checkpoint folders to see what each one holds.

If you work in the field, note that the card gives no architecture details, sizes or benchmark numbers. It points to papers for the two checkpoint sets and to a separate data summary card. Treat it as a package for training your own models, plus a few published checkpoints.

Collect your molecule data
Train a custom model with the package
Compare with the published checkpoints
Apply the model to your molecule task

Good to know: The advanced pre-trained versions are not in this repository.

• License: MIT
• Cost to run: none named for the files here. The card gives no price for the advanced versions on Azure Quantum Elements.
• Data it reaches: it downloads files from Hugging Face to your computer. The card names no account, web or clipboard access.
• Not for: anyone expecting the advanced pre-trained versions here, since they are not in this repository

On Hugging Face, 13 November 2026

• Downloads in the last 30 days: 0
• Likes: 10
• License: mit

Links

• Model: https://huggingface.co/microsoft/Graphormer
• Files: https://huggingface.co/microsoft/Graphormer/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download microsoft/Graphormer

Then follow the usage in its model card. Nothing in this kit runs by itself.