Move safety detection to model call in Flax safety checker #1023
Conversation
This file contains 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
jonatanklosko
commented
Oct 27, 2022
|
The documentation is not available anymore as the PR was closed or merged. |
|
This looks good to me - ok to merge! What do you think @pcuenca ? |
pcuenca
approved these changes
Oct 30, 2022
|
After these changes, I believe it should now be possible to further simplify the code and make it more jax-like by:
Therefore, it'd be up to the user whether to We can address those changes in an upcoming PR :) |
yoonseokjin
pushed a commit
to yoonseokjin/diffusers
that referenced
this pull request
Dec 25, 2023
…ce#1023) * Move safety detection to model call in Flax safety checker * Update src/diffusers/pipelines/stable_diffusion/safety_checker_flax.py
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
Currently a part of the safety checker is outside the model forward pass and is run separately without jitting. This PR merges it into the model call by vectorizing the relevant loops.
See #558 (comment).
I tested that the concept scores before and after this change are the same, and that 2 unsafe among 4 total images are still properly detected.