Learning a Mini-batch Graph Transformer via Two-stage Interaction Augmentation

Official codebase for paper Learning a Mini-batch Graph Transformer via Two-stage Interaction Augmentation.

1. Prerequisites

Install dependencies

See requirment.txt file for more information about how to install the dependencies.

2. Run the code

We provide scripts to replicate the results in the paper.

sh run.sh

Additional Information: Time Comparison

Table R1: The running times for large-scale datasets were recorded. The reported times represent the model training time for a single epoch, measured in seconds.

Methods	ogbn-arxiv	pokec	twitch-gamer
DIFFormer	0.403	4.121	0.595
NodeFormer	0.989	12.827	1.189
NAGphormer	1.857	17.460	1.798
GOAT	13.523	628.67	92.55
LGMformer	24.746	157.419	58.202