How to enable reduction parallel in substitutions?

[weilinquan]

When I use flexflow to search parallel strategy, I found there is only partition_linear_combine. I think it represents data parallel. But I can't find any rules, such as partition_linear_reduce to represents tensor parallel. Should I do something more? Thank you for your help.