I have a question about distributed learning on pytorch geometric.

[Cow-Kite]

I'm currently trying out the new package distributed in pytorch geometric.
I am running the /examples/distributed/pyg example posted on github.
I have a question. Can I only create one trainer process? Or can I create multiple trainers on one node to further accelerate distributed learning?
thank you.

[rusty1s]

Do you mean you want to run multiple processes of DistributedDataParallel on the same node? This is generally possible by simple executing the same script on the same node twice. Does that answer your question?

[Cow-Kite]

Thank you very much!