Running issue of LINKX model

[VeritasYin]

Hello,

I got an issue of running LINKX model for scalable experiments. In particular, in the forward function of LINKX model, it needs to construct a sparse tensor (line 39, models.py), whose dimension is set to (m, self.num_nodes). m is the number of unique nodes in the parsed data batch. However, since the edge indices are not relabeled, there are certainly some node indices exceeding the number of unique nodes (m value here). Is this a bug or I got some wrong here?

Thanks!