Do the optimizations used instead of xformers have an impact on Textual Inversion (and Custom Diffusion) training?

[sebastiandoesart]

I'm enjoying using your fork. Thank your efforts!
For training though, I use the normal Automatic installation with xformers not enabled.
I would use this version but I read that xformers have a negative impact on training, and have no idea if this applies to the optimizations that your fork uses instead of xformers.
I'd prefer to do it in this version though, as moving the embeddings+deltas gets annoying.

Does anybody have any insight into this?
Thanks!