Official codebase for paper Learning a Mini-batch Graph Transformer via Two-stage Interaction Augmentation.
See requirment.txt
file for more information about how to install the dependencies.
We provide scripts to replicate the results in the paper.
sh run.sh
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 |