ENH: Improve speed of function expanding LoRA scales by BenjaminBossan · Pull Request #11834 · huggingface/diffusers (original) (raw)

What does this PR do?

Resolves #11816

The following call proved to be a bottleneck when setting a lot of LoRA adapters in diffusers:

This is because we would repeatedly call unet.state_dict(), even though in the standard case, it is not necessary:

This PR fixes this by deferring this call, so that it is only run when it's necessary, not earlier.

Note: This PR doesn't change the fact that set_adapters becomes slower the more adapters are already loaded, but since it speeds up the whole process by a factor of approximately 10x, set_adapters is much less of a bottleneck.

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline?
• Did you read our philosophy doc (important for complex PRs)?
• Was this discussed/approved via a GitHub issue or the forum? Please add a link to it if that's the case.
• Did you make sure to update the documentation with your changes? Here are the
  documentation guidelines, and
  here are tips on formatting docstrings.
• Did you write any new necessary tests?