fix CI bug for kandinsky3_img2img case #12474
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Signed-off-by: Liu, Kaixuan <kaixuan.liu@intel.com>
|
@a-r-r-o-w @DN6 ， Hi, can you help review? Thx! |
|
@kaixuanliu Does this only affect XPU? Or GPU too? |
|
@regisss This PR also affects GPU, although on GPU, the tolerance value is OK. Steps to reproduce: |
|
@kaixuanliu better solution here is to change the default value to True in the encode_prompt signature since it's hardcoded in the pipeline anyway. |
Signed-off-by: Liu, Kaixuan <kaixuan.liu@intel.com>
|
@DN6 I agree. Have updated the code. |
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
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Outdated suggestions cannot be applied.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.
For full pipeline, related
cut_contextis set toTruein L511 while for pipe_without_text_encoders, this param is set toFalseby default L116, hence there will be some minor difference between their outputs(in UNet L492 the mean value w/ and w/ocut_contextis different). For platforms such as Intel XPU, they use their seperate generator L145, hence the default tolerance value1e-4is not enough. Here we explicitly setencode_prompt_inputs["_cut_context"] = Trueto align with the behavior of full pipeline. @DN6 , pls help review, Thx!