Is it safe to assume that conv layers without a batch argument for forward are safe to use with batched graphs?

[callumm-graphcore]

Hello,

I'm researching PyG because we are planning on writing tutorials on it for my company's platform. I've noticed that some convolutional layers take an optional batch argument to their forward method containing a batch vector. Does this mean that we can assume that the layers that don't allow you to provide such an argument can be safely used with "batched" graphs?

With thanks,
Callum

[rusty1s]

Yes. Any GNN layer without a batch argument can be used for single graphs and batched graphs.

[callumm-graphcore]

Great, thanks Matthias!