# PetarV-/GAT

Graph Attention Networks (https://arxiv.org/abs/1710.10903)

Repository: https://github.com/PetarV-/GAT
Canonical: https://ross.abutalabs.com/products/gat
Homepage: https://petar-v.com/GAT/
Language: Python
License: MIT
License Family: permissive
Topics: graph-attention-networks, attention-mechanism, self-attention, tensorflow, neural-networks, python
Last push: 2022-04-09T20:10:06+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 3136, "days_push": 1607, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 3548, forks 671 (observed 2026-08-28T04:08:09.429077+00:00)

## What it is
Reference implementation of Graph Attention Networks (GAT), the ICLR 2018 model by Veličković et al., written in TensorFlow 1.x with a minimal Cora training example and an experimental sparse variant. It is primarily a research reference rather than a production library.

## Use cases
- implement graph attention networks
- train a GAT model on the Cora citation dataset
- understand how masked self-attention works on graphs
- reproduce results from the GAT paper
- get a reference GAT layer implementation in TensorFlow
- learn graph representation learning from a canonical example

## When to choose
- you need the original reference implementation of GAT for research or citation
- you want a minimal, readable TensorFlow example of attention on graphs
- you are reproducing the ICLR 2018 paper's Cora results

## When to avoid
- you need a maintained, optimized GNN library for production (use PyTorch Geometric, DGL, or Spektral)
- you require modern TensorFlow 2.x or PyTorch support
- you need large-scale or inductive training beyond the toy Cora example

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning
- domain: machine-learning, deep-learning
- platform: python
- tags: graph-attention-networks, graph-neural-networks, tensorflow, attention-mechanism, self-attention, cora, node-classification, research-code, algorithms, gpu

## Member repositories
- PetarV-/GAT (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:08:09.429077+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T18:34:22.818915+00:00, confidence not recorded.
  - readme: https://github.com/PetarV-/GAT (fetched 2026-08-28T04:08:09.429077+00:00, sha ff3d2e9189c2)
  - homepage: https://petar-v.com/GAT/ (fetched 2026-08-29T09:28:26.371205+00:00, sha 1554eadd6b1a)
- Data as of 2026-08-30T08:39:29.467469+00:00.