Backprop through dgl.unbatch?

Hi,

I would like to use unbatch inside a model and then iterate over the batch graphs to compute the final model output. However, I am unsure if backpropagation would still work and if gradients would flow through unbatch. Can someone please clarify if this is ok?

Thanks,

It should be fine. If you have a minimal code snippet to share, you can also post it for a sanity check.