[docs] Caching methods #11625
Merged
[docs] Caching methods #11625
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
|
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. |
a-r-r-o-w
approved these changes
May 29, 2025
docs/source/en/optimization/cache.md
Outdated
|
|
||
| ## FasterCache | ||
|
|
||
| [FasterCache](https://huggingface.co/papers/2410.19355) computes and caches attention features at every other timestep instead of directly reusing cached features because it can cause flickering or blurry details in the generated video. The features from the skipped step are calculated from the difference between the adjacent cached features. |
There was a problem hiding this comment.
Suggested change
| [FasterCache](https://huggingface.co/papers/2410.19355) computes and caches attention features at every other timestep instead of directly reusing cached features because it can cause flickering or blurry details in the generated video. The features from the skipped step are calculated from the difference between the adjacent cached features. | |
| [FasterCache](https://huggingface.co/papers/2410.19355) caches and reuses attention features in a similar manner to PAB, as output differences in successive timesteps of the generation process is small. Additionally, when using classifier-free guidance for sampling (commonly used in most base models), FasterCache may choose to skip the unconditional branch prediction entirely, and estimate it from the conditional branch prediction, if there is a significant redundancy in the predicted latent outputs between successive timesteps. |
There was a problem hiding this comment.
Cc: @sunovivid you might be interested.
sayakpaul
reviewed
May 30, 2025
There was a problem hiding this comment.
Looks good to me. Thanks to @a-r-r-o-w for the suggestions as well. I would be in favor of keeping the technincal details.
Two things:
- Include a table to report timing and memory numbers so that users can know the trade-offs (can happen in a follow-up PR).
- If we have info if the caching methods are generally model-agnostic, having an explicit note about it would be useful.
|
Thanks for the reviews! Happy to include a table in follow-up if someone can provide me with the timing and memory numbers (or the code to generate that)! |
sayakpaul
approved these changes
Jun 2, 2025
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.
Make the caching docs more visible and give a little more context behind the methods.