pytorch_sparse
pytorch_sparse copied to clipboard
Why is there no backward for spspmm in the latest version?
https://github.com/rusty1s/pytorch_sparse/blob/master/torch_sparse/spspmm.py
It was removed in version 0.5.1
since the implementation was kinda slow, but I have plans to bring it back ASAP.
I'm looking forward to seeing this feature soon :)
Any updates on this feature?
Not yet, sorry for the delay.
Hi! Any updates on it? Is there any other efficient way to do spspmm with autograd? I tried using torch.sparse.mm, but didn't manage to do it efficiently. Here I posted a question on StackOverflow about it - maybe someone from here could help?