Project AURIONAI Use-Cases Research & Innovation
University of Cambridge
graphfamily

GAT (Graph Attention Network)

Graph neural network that weights neighbours with attention. Paper by Velickovic et al. (2017).

University of CambridgeUniversity of CambridgeMilaMilaCanada
t-SNE plot of node representations learned by a graph attention network, coloured by class.
Levelfamily
Classgraph
Parametersnot in the notes
Contextnot in the notes
Released30 Oct 2017
Opennessopen source
Runs onnot in the notes
Statusactive

Lineage

successor
GAT (Graph Attention Network) to GATv2: successor, high confidenceGAT (Graph Attention Network)GAT (Graph Attention Netwo…University of Cambridge, Oct 2017GATv2GATv2Technion - Israel Institute of Te…
Technion - Israel Institute of TechnologyGATv2 is a child (successor)high confidencenotes
Technion - Israel Institute of TechnologyGATv2 belongs to this familygraph30 May 2021
Where this comes from
Parent and child relations from model cards, Hugging Face base_model fields and the genealogy tables of the research notes. The tree shows two generations each way, 2 models. Confidence and evidence are the notes' own.

R&D Topics

Graph neural networks and graph transformersproduced by

Identity

Developer
University of Cambridge, Mila
Family
not in the notes
Released
30 Oct 2017 sourcesannouncement page
Weights published
–
Record checked
12 Oct 2026
Where this comes from
Developer and dates from the research notes. Release date is the vendor's public announcement; weights date is when open weights were published.

Architecture and Size

Architecture
GNN Masked self-attention over neighbours
Total parameters
– sources
Active parameters
–
Checkpoint parameters
–
Context length
not in the notes
Input
graph
Output
graph
Where this comes from
Parameter counts as the developer states them. Where the checkpoint holds a different total, both are shown (register A16).

Openness and Access

Openness
open source
Licence
MIT License sources
Access routes
not in the notes
Where this comes from
The licence file governs over registry tags and aggregator labels.

Deployment

Runs on
not in the notes
Footprint
Licence is that of the authors' PetarV-/GAT repository (MIT).
Runtimes
not in the notes
Where this comes from
Deployment tiers and runtimes as named in the notes. Footprint is the notes' figure, not a measurement by AURION.

Indonesia

Indonesian support
not stated
Regional languages
not in the notes
Jakarta hosting
unknown
Languages
not in the notes
Where this comes from
Indonesia fields are carried on every model; unknown is an allowed value.

Benchmark Results

The notes record no benchmark result for this model.

Industry Use Cases

No industry use case links this model.

Sources

  1. arxiv.org/abs/1710.10903 https://arxiv.org/abs/1710.10903
  2. github.com/PetarV-/GAT https://github.com/PetarV-/GAT
  3. GitHub - pyg-team/pytorch_geometric: Graph Neural Network Library for PyTorch github.com https://github.com/pyg-team/pytorch_geometric
  4. pytorch-geometric.readthedocs.io/en/latest/modules/nn.html https://pytorch-geometric.readthedocs.io/en/latest/modules/nn.html
  5. modelzoo.co/model/pygat https://modelzoo.co/model/pygat
  6. huggingface.co/datasets/pwc-archive/methods https://huggingface.co/datasets/pwc-archive/methods
  7. arxiv.org/html/1710.10903 https://arxiv.org/html/1710.10903

To cite this page: Project AURION, AI Use-Cases Research & Innovation, "GAT (Graph Attention Network)", https://aurion.semai.id/models/gat/, records of October 2026.