Fix issue with prompt embeds and latents in SD Cascade Decoder with multiple image embeddings for a single prompt. #7381
+79
−4
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
What does this PR do?
There is a use case with the cascade decoder pipeline where a user might set
num_image_per_prompt
> 1 in the prior pipeline and pass the output to the decoder.This results in a case where you have multiple image embeddings associated with a single prompt. This is causing an issue with the how the batch size is set for latents and prompt embeds in the decoder pipeline. T
Fixes #7377
This PR:
Before submitting
documentation guidelines, and
here are tips on formatting docstrings.
Who can review?
Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.