PyGCL / PyGCL

PyGCL: A PyTorch Library for Graph Contrastive Learning

Home Page:https://PyGCL.readthedocs.io

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

How to run GraphCL with Node classification tasks?

zhangtanma opened this issue · comments

Dear authors, this is a great job in Graph Contrastive learning field. But I am encountering 2 small problems.

  1. When I run GRACE with large datasets such as WikiCS, then the CUDA out of memory problem occurs.
  2. How to run GraphCL on node classification problem?
    I am much thankful if you can reply to me !

Hi! Thanks for your interest on our work. For large-scale datasets, you can slightly lower the hidden dimension or try negative-sample-free models such as BGRL and GBT. These two models plus our GRACE all target at node classification and can be found in the examples folder.