Skip to content

Unet2D model use ConfigMixin & ModelMixin #10

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Conversation

mishig25
Copy link

@mishig25 mishig25 commented Sep 11, 2022

Since diffusers models use ConfigMixin rather than Config Class, I've made:

  1. Unet2D use ConfigMixin & removed the Unet2DConfig class entirely
  2. Unet2D use ModelMixin, which is mostly copy of transformers.FlaxPretrainedModel but without the functionalities that are needed currently in diffusers (i.e. I've removed functionalities such as sharded checkpoints, pt-flax checkpint conversion, and other features that are not found in diffusers.ModelMixin currently)

Please let me know how this PR is looking 😊 If it is not a change we want, happy to just close it. Otherwise, happy to extend the changes to AutoencoderKL model & add necessary tests

cc: @patil-suraj @patrickvonplaten @pcuenca (I guess because of permissions can't add you as reviewers)

@patil-suraj
Copy link
Owner

patil-suraj commented Sep 12, 2022

Super cool Mishig! The changes look good!
This will now officially become part of diffusers, so let's maybe integrate it there :)

@mishig25
Copy link
Author

oh I see, please let me know what should be the plan.

In diffusers repo, do we plan to have:

  1. create branch jax_main_branch
  2. create/merge various PRs into jax_main_branch
  3. merge jax_main_branch into main

or will the process be something different?

@mishig25 mishig25 closed this Sep 20, 2022
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants