Allow DDPM scheduler to use model's predicated variance #132
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
to utilize model that also predict the variance.
anton-l
approved these changes
Aug 3, 2022
There was a problem hiding this comment.
Hi @eyalmazuz! Thank you for your contribution, this will be quite useful for when we bring OpenAI's guided diffusion and Glide models into the library!
We'll have to think a bit more about the API design for learned variance (cc @patil-suraj), but for now this will work fine :)
|
The documentation is not available anymore as the PR was closed or merged. |
yoonseokjin
pushed a commit
to yoonseokjin/diffusers
that referenced
this pull request
Dec 25, 2023
) * Extented the ability of ddpm scheduler to utilize model that also predict the variance. * Update src/diffusers/schedulers/scheduling_ddpm.py Co-authored-by: Anton Lozhkov <[email protected]>
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Outdated suggestions cannot be applied.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.
This is an implementation of the Improved Denoising Diffusion Probabilistic Models
which added the ability to learn the variance instead of just predicting the noise.
It is assumed that the step method receives a tensor shape (B, C, ...) when predicting noise.
At the moment, it only implements variance. In order to use it fully, the user needs to write the additional loss function from the paper into their training loop.
For future contribution, I think it would be beneficial to implement a losses module with different loss functions like the VB loss in the Improved Diffusion paper or other loss functions used in different diffusion papers (and maybe a trainer object). If needed, I can make new PRs with the relevant additions, or this can be an ongoing PR (the former seems better in case I'll stop contributing for some reason).