Why does reduce_sum_p takes out_sharding as an argument, but reduce_max_p and reduce_min_p do not?

[botev]

The title.

[yashk2810]

No reason. Can you explain your use case and give a minimal repro? I can add it then.

[botev]

I was writing code that is trying to produce the same thing for different reduction primitives, and I hit this issue, because the reduce_sum takes different argument. I was wondering if there is any specific reason for this, or is it just not having enough time to add it everywhere.

[yashk2810]

The latter. I can add it though if you want.

[botev]

I would appreciate if the interface is indeed the same.