关于GCU、沐曦GPGPU、MLU、0卡V100资源4月7日恢复上架的公告>>> 关于共建具身智能开源数据集的倡议>>> 关于云脑任务中统一路径访问方式的公告>>> 关于将启智集群GPU资源迁移至智算集群的公告>>>
zhanghongyi
/
dgl
mirror of https://github.com/dmlc/dgl
Not watched
1
0
Fork 2
Code
Releases 40 Wiki Activity
Issues 0 Datasets Model Cloudbrain
You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.
4103 Commits
27 Branches
329 MiB
166 Download
 
 
 
 
 
 
Branch: master
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'master'
${ noResults }
dgl/examples/pytorch/hardgat
Andrei Ivanov 0c3b2b780f
Improving the HARD_GAT example (#6036) 		9 months ago
..
README.md [Example] Add hardGAT example (#2486) 3 years ago
hgao.py examples (#5323) 1 year ago
train.py Improving the HARD_GAT example (#6036) 9 months ago
utils.py [Misc] Black auto fix. (#4651) 1 year ago

README.md

HardGAT

DGL Implementation of h/cGAO paper.

This DGL example implements the GNN model proposed in the paper HardGraphAttention.

HardGANet implementor

This example was implemented by Ericcsr during his Internship work at the AWS Shanghai AI Lab.

The graph dataset used in this example

The DGL's built-in CoraGraphDataset. Dataset summary:

  • NumNodes: 2708
  • NumEdges: 10556
  • NumFeats: 1433
  • NumClasses: 7
  • NumTrainingSamples: 140
  • NumValidationSamples: 500
  • NumTestSamples: 1000

The DGL's build-in CiteseerGraphDataset. Dataset Summary:

  • NumNodes: 3327
  • NumEdges: 9228
  • NumFeats: 3703
  • NumClasses: 6
  • NumTrainingSamples: 120
  • NumValidationSamples: 500
  • NumTestSamples: 1000

The DGL's build-in PubmedGraphDataset. Dataset Summary:

  • NumNodes: 19717
  • NumEdges: 88651
  • NumFeats: 500
  • NumClasses: 3
  • NumTrainingSamples: 60
  • NumValidationSamples: 500
  • NumTestSamples: 1000

How to run example files

In the hgao folder, run

Please use train.py

python train.py --dataset=cora

If want to use a GPU, run

python train.py --gpu 0 --dataset=citeseer

If you want to use more Graph Hard Attention Modules

python train.py --num-layers <your number> --dataset=pubmed

If you want to change the hard attention threshold k

python train.py --k <your number> --dataset=cora

If you want to test with vanillia GAT

python train.py --model <gat/hgat> --dataset=cora

Performance

Models/Datasets Cora Citeseer Pubmed
GAT in DGL 81.5% 70.1% 77.7%
HardGAT 81.8% 70.2% 78.0%

Notice that HardGAT Simply replace GATConv with hGAO mentioned in paper.

Delete task

Are you sure you want to delete this task? Once this task is deleted, it cannot be recovered.

cancel
confirm

No Description

Python C++ Jupyter Notebook Cuda Text other

Please read the following content carefully:

Dear OpenI User

Thank you for your continuous support to the Openl Qizhi Community AI Collaboration Platform. In order to protect your usage rights and ensure network security, we updated the Openl Qizhi Community AI Collaboration Platform Usage Agreement in January 2024. The updated agreement specifies that users are prohibited from using intranet penetration tools. After you click "Agree and continue", you can continue to use our services. Thank you for your cooperation and understanding.

For more agreement content, please refer to the《Openl Qizhi Community AI Collaboration Platform Usage Agreement》

Agree and continue
Disagree, exit