I have two sparse tensors a
and b
of sizes (A, B, C, D)
and (A, B)
respectively.
How do I multiply the two of them?
I don’t want to copy b
across a new third and fourth dimension C
and D
times because that wastes memory and also takes compute time. I also don’t want to convert both to dense and back to sparse (although converting b
to dense at some point would be fine; it’s just a
that takes time to convert).