Support OLSS Scheduler

[ZihaoW123]

OLSS is the most advanced sampler. Can you consider adding it to the scheduler library of diffusers
(https://github.com/alibaba/EasyNLP/tree/master/diffusion/olss_scheduler)

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

Please note that issues that do not follow the contributing guidelines are likely to be ignored.

[a-r-r-o-w]

cc @yiyixuxu @asomoza @vladmandic

[vladmandic]

i took a look at olss a while back, wasn't really sure how to use it
(i don't mean technically, i mean from the workflow perspective).

it requires (significant) training before it can be used.
simplified, its like runing ddim for steps=100 so later olss works at lower steps.
if model or prompt change significantly, it needs to be retrained - and all that makes scheduler stateful instead of stateless.
and in my workflow, i don't maintain scheduler state, its instantiated as-needed.
i'm sure it could be wrapped with additional layer that does all that behind the scenes, but i didn't see sufficient interest for that.

btw, another advanced scheduler with similar pre-train approach is https://github.com/wl-zhao/DC-Solver

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

Please note that issues that do not follow the contributing guidelines are likely to be ignored.