MMDiT implementation and text-to-image training with rectified flows #155
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
A-Jacobson
reviewed
Jul 9, 2024
A-Jacobson
approved these changes
Jul 9, 2024
There was a problem hiding this comment.
Left a few commends about structure. Basically, I'd like to see more of the transformer logic confined to the transformer with the ComposerModel getting as close to "create the 3 models and call them" as possible. Overall, it's awesome and I conditionally approve as it's non-breaking and also has successful test runs.
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Outdated suggestions cannot be applied.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 PR contains an implementation of the MMDiT model from the SD3 paper, along with a model class for using it to train text to image models. To support this, a generic model inference class is also included.
Major additions:
diffusion/inference/inference_model.pyhasModelnferenceclass for inference with arbitrary models frommodels.pydiffusion/models/models.pyIncludes atext_to_image_transformermodel for SD3 style MMDiTdiffusion/models/t2i_transformer.pyhas theComposerModelclass for the MMDiT text to image modeldiffusion/models/transformer.pyhas the layers/blocks for the MMDiT modeldiffusion/train.pyincludes a new function to configure the optimizer for the new text to image model