Libtorch sparse x dense batch multiplication

Hi,

Is there any way to perform broadcasting matrix product between sparse A[1x13x13] and dense B[1024x13x139]?
If A is dense torch::matmul works fine, but if it’s sparse everything crashes.
I’m using libtorch 1.7.0
If it’s impossible right now then probably there’s some workaround?

3 Likes