diegoantognini/pyGAT：Pytorch implementation of the

diegoantognini

3.1k

700 47

MIT License

项目介绍

diegoantognini/pyGAT 是一个托管在 GitHub 上的开源项目，主要使用 Python 编写，采用 MIT License 许可证，目前获得 3.1k 个 Star、700 次 Fork。

仓库全名: diegoantognini/pyGAT
开源许可证: MIT License
Star 数: 3.1k
Fork 数: 700
未关闭 Issue: 47

README 摘要

Pytorch Graph Attention Network This is a pytorch implementation of the Graph Attention Network (GAT) model presented by Veličković et. al (2017, https://arxiv.org/abs/1710.10903). The repo has been forked initially from https://github.com/tkipf/pygcn. The official repository for the GAT (Tensorflow) is available in https://github.com/PetarV-/GAT. Therefore, if you make advantage of the pyGAT model in your research,…

完整内容见 GitHub 原始 README 或本页。

常见问题

pyGAT 用什么语言开发？

主要使用 Python 编写，采用 MIT License 许可证，目前有 3.1k 个 Star。

pyGAT 可以免费商用吗？

如何获取 pyGAT 的源码？

可以直接克隆仓库：git clone https://github.com/diegoantognini/pyGAT.git，默认分支为 master；安装与运行步骤见仓库 README。

pyGAT 最近还在维护吗？

仓库最近一次更新为 2023年7月7日，当前有 47 个未关闭 Issue；Star 数为 3.1k，可作为社区活跃度的参考。

简要分析

热度分

4536

价值分

2082

活跃状态

活跃

attention-mechanism graph-attention-networks neural-networks python pytorch self-attention

GitHub

语言占比

Python

贡献者

README

Pytorch Graph Attention Network

This is a pytorch implementation of the Graph Attention Network (GAT) model presented by Veličković et. al (2017, https://arxiv.org/abs/1710.10903).

The repo has been forked initially from https://github.com/tkipf/pygcn. The official repository for the GAT (Tensorflow) is available in https://github.com/PetarV-/GAT. Therefore, if you make advantage of the pyGAT model in your research, please cite the following:

@article{
  velickovic2018graph,
  title="{Graph Attention Networks}",
  author={Veli{\v{c}}kovi{\'{c}}, Petar and Cucurull, Guillem and Casanova, Arantxa and Romero, Adriana and Li{\`{o}}, Pietro and Bengio, Yoshua},
  journal={International Conference on Learning Representations},
  year={2018},
  url={https://openreview.net/forum?id=rJXMpikCZ},
  note={accepted as poster},
}

The branch master contains the implementation from the paper. The branch similar_impl_tensorflow the implementation from the official Tensorflow repository.

Performances

For the branch master, the training of the transductive learning on Cora task on a Titan Xp takes ~0.9 sec per epoch and 10-15 minutes for the whole training (~800 epochs). The final accuracy is between 84.2 and 85.3 (obtained on 5 different runs). For the branch similar_impl_tensorflow, the training takes less than 1 minute and reach ~83.0.

A small note about initial sparse matrix operations of https://github.com/tkipf/pygcn: they have been removed. Therefore, the current model take ~7GB on GRAM.

Sparse version GAT

We develop a sparse version GAT using pytorch. There are numerically instability because of softmax function. Therefore, you need to initialize carefully. To use sparse version GAT, add flag --sparse. The performance of sparse version is similar with tensorflow. On a Titan Xp takes 0.08~0.14 sec.

Requirements

pyGAT relies on Python 3.5 and PyTorch 0.4.1 (due to torch.sparse_coo_tensor).

Issues/Pull Requests/Feedbacks

Don’t hesitate to contact for any feedback or create issues/pull requests.

pyGAT