Sparse Bilinear map

[Xparx]

Hey,

I'm trying to implement a custom attention mechanism.
Is there a function or a efficient way to implement a bilinear mapping in pytorch_sparse?

[rusty1s]

Can you clarify what you wan to do exactly?

[Xparx]

I wanted to try to do the same as in the pytorch bilinear module, when A is a sparse matrix.

[Xparx]

As far as I understand pytorch_sparse works on 2d tensors so might not be relevant exactly. I might need to think this one over a bit.

[rusty1s]

torch-sparse can handle more than 2 dimensions (but only the first two can be sparse). Would that be sufficient in your case?