Description
torch-geometric is an agent skill for graph neural networks with PyTorch Geometric, the standard library for graph learning built on PyTorch. It covers node, link and graph classification, message-passing layers such as GCN, GAT, GraphSAGE and GIN, heterogeneous graphs, and scalable mini-batch training. The library provides graph data structures and more than 60 GNN layer implementations. The skill reviews torch-geometric 2.8.0.post1, released 1 October 2026, and its CPU examples were tested with Python 3.13 and PyTorch 2.14.1. The rolling documentation identifies 2.9.0, so the skill tells you to check the installed release before using a newer API. PyG 2.8 requires PyTorch 2.9 or newer. Basic tensor-based layers need no extensions, but neighbour sampling needs pyg-lib or torch-sparse, and spatial nearest-neighbour operators need pyg-lib in 2.8. The torch-cluster and torch-spline-conv packages are deprecated and ignored. Install the PyTorch build for your platform first, and never install wheels for a different PyTorch release just because core imports work. Use it when your data is naturally a graph, such as molecules, interaction networks or patient similarity networks.…
Highlights
- Node, link and graph classification
- GCN, GAT, GraphSAGE and GIN layers
- Heterogeneous graphs
- Mini-batch training at scale
- Reviewed on torch-geometric 2.8.0.post1
Ratings & Reviews
No ratings yet
Be the first to rate torch-geometric.
People also visit
pytorch-patterns
PyTorch deep learning patterns and best practices for building robust
torch-geometric FAQ
What does PyTorch Geometric provide?
Graph data structures, more than 60 GNN layers, mini-batch training and heterogeneous graph support.
Which version was reviewed?
torch-geometric 2.8.0.post1, reviewed on 1 October 2026, with CPU tests on Python 3.13 and PyTorch 2.14.1.
Which PyTorch version does it need?
PyG 2.8 requires PyTorch 2.9 or newer.
Which extensions are deprecated?
torch-cluster and torch-spline-conv. The skill says they are ignored.
How do I install it?
Run npx skills add https://github.com/k-dense-ai/scientific-agent-skills --skill torch-geometric.
Is it checked for security?
Its security audits show passes from Gen Agent Trust Hub, Socket and Snyk.
What are the best torch-geometric alternatives?
Popular alternatives to torch-geometric include pytorch-patterns, aeon and esm. See all torch-geometric alternatives