caojy1998

> Hi @caojy1998, thanks for reporting this issue. I have encountered this issue before while installing all these dependencies via an environment yaml file. Since `setup.py` imports `torch` at: https://github.com/rusty1s/pytorch_scatter/blob/32caa1c624f164460788b352796f3022fe9a4d72/setup.py#L8...

> @caojy1998 do you have any timeline on when we can have `ogbn-papers100M` as a built-in dataset? I would like to update the multiGPU example to have the papers100M as...

Currently Papers100M is added.

I have changed the indice dtype to int32.

> > I have changed the indice dtype to int32. > > Now, we are getting an error because train_set etc. is not in int32. Will need #7127 to be...

> I can quickly handle the coding part, but I might need assistance from @caojy1998 as I lack certain permissions. Previously, a lot of misaligned (benchmark) settings resulted in this...