mmdit

Implementation of a single layer of the MMDiT, proposed in Stable Diffusion 3, in Pytorch

Project.

Catalog snapshot

License: MIT
Platforms: Linux
Docker: No
Commercial use: Allowed
Stars: 559
Language: Python
Last commit: Jan 18, 2026
Tags: artificial-intelligenceattention-mechanismsdeep-learningmulti-modal-attention

Repository health snapshot: 17 forks and 3 open issues at review time. Platform support is inferred conservatively from repository topics.

About.

Implementation of a single layer of the MMDiT, proposed in Stable Diffusion 3, in Pytorch Fossfits includes mmdit as an actively maintained image ai option: its public repository was updated on 2026-01-18, uses the MIT license, and had 559 stars when reviewed. Check its documentation and release notes before production adoption.

artificial intelligence · attention mechanisms · deep learning · multi modal attention

Nearby.

All Image AI

SDStable Diffusion web UIThe AUTOMATIC1111 Gradio UI that made local SD mainstream.Compare
Sstable-diffusion.cppPure C/C++ inference for SD, Flux, and Qwen-Image weights — CPU included.Compare
FFooocusMidjourney-like simplicity on top of local diffusion.Compare
KAKrita AI DiffusionKrita plugin that generates, inpaints, and outpaints as native layers via ComfyUI.Compare
IInvokeAIPolished local creative tool around diffusion, with a studio UI.Compare