fix(ddim): validate eta is in [0, 1] in DDIMPipeline#13367
Merged
yiyixuxu merged 4 commits intohuggingface:mainfrom Apr 3, 2026
Merged
fix(ddim): validate eta is in [0, 1] in DDIMPipeline#13367yiyixuxu merged 4 commits intohuggingface:mainfrom
yiyixuxu merged 4 commits intohuggingface:mainfrom
Conversation
The DDIM paper defines η (eta) as a value that must lie in [0, 1]: η=0 corresponds to deterministic DDIM, η=1 corresponds to DDPM. The docstring already documented this constraint, but no runtime validation was in place, so users could silently pass out-of-range values (e.g. negative or >1) without any error. Add an explicit ValueError check before the denoising loop so that invalid eta values are caught early with a clear message. Fixes huggingface#13362 Signed-off-by: NIK-TIGER-BILL <nik.tiger.bill@github.com>
|
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. |
Collaborator
|
thanks for the PR, can you change it to a warning? |
Per maintainer feedback from @yiyixuxu — the documentation is sufficient; a hard ValueError is too strict. Replace with a UserWarning so callers are informed without breaking existing code that passes eta outside [0, 1]. Signed-off-by: NIK-TIGER-BILL <nik.tiger.bill@github.com>
Contributor
Author
|
Thanks for the feedback, @yiyixuxu! Changed to a |
yiyixuxu
reviewed
Mar 31, 2026
| image_shape = (batch_size, self.unet.config.in_channels, *self.unet.config.sample_size) | ||
|
|
||
| if not 0.0 <= eta <= 1.0: | ||
| warnings.warn( |
yiyixuxu
reviewed
Mar 31, 2026
| image_shape = (batch_size, self.unet.config.in_channels, *self.unet.config.sample_size) | ||
|
|
||
| if not 0.0 <= eta <= 1.0: | ||
| warnings.warn( |
Collaborator
There was a problem hiding this comment.
can we change it to use logger.warning? to be consistent with all other pipelines
Contributor
Author
There was a problem hiding this comment.
Good call! Updated in the latest commit — switched from warnings.warn() to logger.warning() to match the convention used in all other diffusers pipelines. Also removed the now-unused import warnings.
…tion Address review request from @yiyixuxu: switch from warnings.warn() to logger.warning() to be consistent with all other diffusers pipelines. The eta validation check itself (0.0 <= eta <= 1.0) is unchanged. Signed-off-by: NIK-TIGER-BILL <nik.tiger.bill@github.com>
terarachang
pushed a commit
to terarachang/diffusers
that referenced
this pull request
Apr 30, 2026
* fix(ddim): validate eta is in [0, 1] in DDIMPipeline.__call__ The DDIM paper defines η (eta) as a value that must lie in [0, 1]: η=0 corresponds to deterministic DDIM, η=1 corresponds to DDPM. The docstring already documented this constraint, but no runtime validation was in place, so users could silently pass out-of-range values (e.g. negative or >1) without any error. Add an explicit ValueError check before the denoising loop so that invalid eta values are caught early with a clear message. Fixes huggingface#13362 Signed-off-by: NIK-TIGER-BILL <nik.tiger.bill@github.com> * fix(ddim): downgrade eta out-of-range from error to warning Per maintainer feedback from @yiyixuxu — the documentation is sufficient; a hard ValueError is too strict. Replace with a UserWarning so callers are informed without breaking existing code that passes eta outside [0, 1]. Signed-off-by: NIK-TIGER-BILL <nik.tiger.bill@github.com> * fix(ddim): use logger.warning instead of warnings.warn for eta validation Address review request from @yiyixuxu: switch from warnings.warn() to logger.warning() to be consistent with all other diffusers pipelines. The eta validation check itself (0.0 <= eta <= 1.0) is unchanged. Signed-off-by: NIK-TIGER-BILL <nik.tiger.bill@github.com> --------- Signed-off-by: NIK-TIGER-BILL <nik.tiger.bill@github.com> Co-authored-by: NIK-TIGER-BILL <nik.tiger.bill@github.com> Co-authored-by: YiYi Xu <yixu310@gmail.com>
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 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.
What does this PR do?
Adds input validation for the
etaparameter inDDIMPipeline.__call__.Fixes #13362
Context
The DDIM paper defines η (eta) as a value that lies strictly in [0, 1]:
The docstring already documents this constraint:
However, no runtime check existed, so out-of-range values (negative or > 1) were silently accepted and could produce undefined/unintended sampling behaviour.
Change
Add a
ValueErrorguard immediately after the image-shape computation, before the denoising loop:Before submitting
Who can review?
@pcuenca @DN6