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,