SDV
SDV copied to clipboard
Dependency Versioning Stategy
Problem Description
Currently, all libraries in the SDV ecosystem have narrow and fixed dependency ranges (with a min and max).
Sometimes as external libraries change, users may have problems installing the SDV such as #1349, #1360, #1345.
Expected behavior
I'm filing this as a broad issue to think about the SDV dependency versioning strategy. In addition to supporting the latest version of pandas (#1366) and pytorch (#1365), we should also think about how to enable the broadest set of of acceptable versions.