haidog-yaqub/MeanFlow
Pytorch Implementation (unofficial) of the paper "Mean Flows for One-step Generative Modeling" by Geng et al.
Implements mean flow-based one-step generation using Jacobian-vector products (JVP) for direct sample prediction from noise, supporting classifier-free guidance and latent representations. Built on DiT architecture with multi-GPU training via Hugging Face Accelerate, though PyTorch's JVP creates GPU memory overhead and incompatibilities with optimizations like Flash Attention. The codebase intentionally omits logging frameworks to maintain minimal dependencies for easy project integration.
1,093 stars.
Stars
1,093
Forks
60
Language
Python
License
MIT
Category
Last pushed
Dec 17, 2025
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/diffusion/haidog-yaqub/MeanFlow"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
milad1378yz/MOTFM
Flow Matching for Medical Image Synthesis: Bridging the Gap Between Speed and Quality
X-GenGroup/Flow-Factory
A unified framework for easy reinforcement learning in Flow-Matching models
OpenImagingLab/FlashVSR
[CVPR 2026] Towards Real-Time Diffusion-Based Streaming Video Super-Resolution — An efficient...
fallenshock/FlowEdit
Official implementation of the paper: "FlowEdit: Inversion-Free Text-Based Editing Using...
jy0205/Pyramid-Flow
[ICLR 2025] Pyramidal Flow Matching for Efficient Video Generative Modeling