qiaoyu-tan / sugrl

SUGRL: Simple Unsupervised Graph Representation Learning

This repository contains the reference code for the paper Simple Unsupervised Graph Representation Learning

Contents

0. Installation
1. Preparation
2. Training
3. Testing

Installation

pip install -r requirements.txt

Preparation

Pretrained model see >>>here<<<.

Configs see >>>here<<<.

Dataset (--dataset-class, --dataset-name,--Custom-key)

Dataset class	Dataset name	Custom key
Planetoid	Cora	classification
Planetoid	CiteSeer	classification
Planetoid	PubMed	classification
MyAmazon	Photo	classification
MyAmazon	Computers	classification
PygNodePropPredDataset	ogbn-arxiv	classification
PygNodePropPredDataset	ogbn-mag	classification
PygNodePropPredDataset	ogbn-products	classification

Important args:

• --pretrain Test checkpoints
• --dataset-class Planetoid, MyAmazon, PygNodePropPredDataset
• --dataset-name Cora, CiteSeer, PubMed, Photo, Computers, ogbn-arxiv, ogbn-mag, ogbn-products
• --custom_key classification, link, clu

Training

python train.py 

Testing

Choose the custom_key of different downstream tasks