OGB and Reddit experiments

[YcZ76]

Hi, allen，thanks for your promising work in this paper. As you say, these two datasets are large. Cause the scalability is under my consideration when i try to explore a new GNN, could you release the code of tackling these datasets? best wishes~~

[allenhaozhu]

Oh, I guess it is not a good time right now. DDL for NIPS video recording, ICLR reviews, CVPR......I guess you only need a svd version of COLES right？ if it like that I can share you the code, it is very simple.

[YcZ76]

thanks~ , btw i just glanced this code. in the model.py, it seems u modify the GRACE's infonce loss with a simplied vesion? however, this model.py seems not to be imported by the others? thanks for your kindly answers, wish u get good news in cvpr/iclr, as well as the camera ready in nips. ;)

[allenhaozhu]

I don't use that. You will find the loss is in every train_**.py.

[YcZ76]

Yep, i see, thanks ~~