Graphormer scalability of edges

[brovatten]

Hi,

I'm using Graphormers on a large graph dataset for pretraining and am currently preprocessing subgraphs for each node. But before I do that, I would be interested to hear how Graphormers scales with regards to edges. In the paper, they only mention it scales quadratically to nodes as a result of attention. Moreover, in the Huggingface documentation they say it typically cannot handle more than $100$ nodes/edges.

Thanks in advance.