Support for asymmetric data matrix?

When I use cuopt.DataModelView.set_matrix on asymmetric distance matrix, I get the following error.

RuntimeError: NVIDIA cuOpt failure - Limited version: Expects data matrix to be symmetric

Is asymmetric distance matrix not supported yet? If so, are there any workarounds?


The Early Access version of CuOpt is designed to support symmetric matrices only. Any non-symmetric matrix can be symmetrized for test purpose by computing A+A^T.