Strided to batch BSR/BSC conversion fails when the number of zeros per block varies while the number of blocks per patch is constant #98495
Labels
module: sparse
Related to torch.sparse
triaged
This issue has been looked at a team member, and triaged and prioritized into an appropriate module
Projects
Issue description
As in the title.
Code example
Notice that in the failing conversion example, the number of zeros in the first block is 0 and in the second block it is 1.
Apparently, the check logic in
pytorch/aten/src/ATen/native/TensorConversions.cpp
Lines 95 to 98 in ccc27bc
is flawed for BSR and BSC conversion cases.
System Info
cc @alexsamardzic @nikitaved @cpuhrsch @amjames @bhosmer
The text was updated successfully, but these errors were encountered: