pyg-team/pytorch_geometric

Graph Positional and Structural Encoder

Open

#8,310 opened on Nov 1, 2023

View on GitHub
 (21 comments) (0 reactions) (0 assignees)Python (19,985 stars) (3,514 forks)batch import
featuregood first issuehelp wantednn

Description

🚀 The feature, motivation and pitch

It would be great to add the "GPSE" (Graph Positional and Structural Encoder) module to PyG, which acts as a trained encoder for Graph Transformers. Most importantly, we should support loading pre-trained encoders and provide them to the user, e.g., via

encoder = GPSE.from_pretrained("molpcba")

Code: https://github.com/G-Taxonomy-Workgroup/GPSE Paper: https://arxiv.org/abs/2307.07107 Twitter: https://twitter.com/RemyLau3/status/1717194030986190899

Any help on this is highly appreciated :)

Alternatives

No response

Additional context

No response

Contributor guide

Tech stack
pythonpytorch
Domain
machine learning
Issue type
feature
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
5
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
over 1 week
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
active
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
clear
Prerequisites
PyTorchPyG basicsgraph neural networks
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
15
Research direction
Review the GPSE repository at https://github.com/G Taxonomy Workgroup/GPSE and the paper for architecture details. Look at existing encoder implementations in PyG (e.g., torch geometric.nn.models) for structure. Implement the GPSE module with a from pretrained method that loads pre trained weights. Ensure integration with PyG's data flow.