Question about Flux attention implementation

Hi,

When I read the code at https://github.com/kohya-ss/sd-scripts/blob/5dff02a65da38c411ee679821504ce947d2abd7d/library/sd3_models.py#L545-L565

I see 3 different optimization approach for attention block in sd3.

But for Flux model, which also have similar MMDiT block, I see attention implementation is https://github.com/kohya-ss/sd-scripts/blob/5dff02a65da38c411ee679821504ce947d2abd7d/library/flux_models.py#L449-L455

which only contains SDPA approach, no xformers approach.

I search for keywords like "xformers" and "flux", but it seems like no one talk about this difference.

So, can I ask the reason behind it? In my opinion, same structure can benefit from same optimization approach. If it is possible to add xformers for flux attention?

Compare the SD3 attention implementation in library/sd3_models.py at lines 545-565 with the Flux implementation in library/flux_models.py at lines 449-455. First determine why Flux uses only SDPA, then assess whether the same xformers optimization can apply and document the conclusion or scope of a change.

Written by the indexing model from the issue text.