Details
-
Bug
-
Status: Closed
-
Major
-
Resolution: Fixed
-
None
-
None
-
None
Description
There are instances of binary operations where the input sizes (e.g., matrix-vector) do not cover a potential dense-sparse change, which causes outputs to remain in wrong sparse/dense representations. For execution types such as singlenode as used in JMLC scenarios this is unnecessary.
Furthermore, for special cases such as ctable rexpand, the output representation cannot be determined upfront and hence, we should similar to binary operations, check for sparse/dense conversions accordingly.