Introducing Graphormer: A Deep Learning Tool for Molecule Science
Graphormer is a comprehensive deep learning package, designed to assist researchers and developers in the field of molecule modeling. It holds the potential to accelerate advancements in AI applications, particularly in the area of material discovery and drug discovery. To learn more about this innovative project, you can visit the official project website.
Advanced Pre-Trained Models
For those interested in utilizing advanced pre-trained versions of Graphormer, these are available exclusively on Azure Quantum Elements. This offers users an edge in the fast-paced and competitive world of AI for molecular science.
Highlights in Graphormer v2.0
Graphormer has introduced several exciting features and updates in its v2.0 release. Some key highlights include:
- Availability of the model, code, and script used in the Open Catalyst Challenge.
- Newly available pre-trained models on PCQM4M and PCQM4Mv2, with additional models on the way.
- Compatibility with various interfaces and datasets, including PyG, DGL, OGB, and OCP.
- Incorporation of the fairseq backbone, enhancing functionality.
- Comprehensive documentation accessible online, facilitating easier use.
Recent Developments and Achievements
Graphormer has made notable strides and accomplishments in the field, including:
- 03/10/2022: A technical report was uploaded, detailing improved benchmarks on PCQM4M and the Open Catalyst Project.
- 12/22/2021: Release of Graphormer v2.0, marking an important milestone for users.
- 12/10/2021: Victory in the Open Catalyst Challenge; technical insights shared in a YouTube talk.
- 09/30/2021: Acceptance by NeurIPS 2021, celebrating Graphormer's innovative contributions.
- 06/16/2021: Secured 1st place in the quantum prediction track of the Open Graph Benchmark Large-Scale Challenge (KDD CUP 2021).
Getting Started with Graphormer
For those looking to dive into Graphormer, comprehensive documentation is available at Graphormer Docs. This repository offers guidance for getting started, training new models, and extending Graphormer with novel model types and tasks. For practical demonstrations, users can explore examples showcasing common task executions via command line.
Requirements and Installation
Setting up Graphormer can be done seamlessly using Conda with a simple installation script:
bash install.sh
Contribution and Community
Graphormer welcomes contributions and suggestions from the community. Interested individuals are encouraged to participate, keeping in mind the requirement to agree to a Contributor License Agreement (CLA). For more details, potential contributors can visit CLA Information. Additionally, adherence to the Microsoft Open Source Code of Conduct ensures a respectful and collaborative environment.
Trademark Notice
The project may include trademarks or logos associated with Microsoft or other entities, each subject to their respective policies. Users are advised to adhere to Microsoft's Trademark & Brand Guidelines when using Microsoft-related assets.
Graphormer stands out as a valuable tool for anyone involved in the realm of molecular science, offering a powerful platform for research and development in AI-driven applications.