gnobitab/RectifiedFlow

Health v2 · maintenance only

Flags: no_releases no_license

Adoption not part of the score

1644 stars · 102 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

Official PyTorch implementation of Rectified Flow, an ICLR 2023 Spotlight method for learning transport maps between distributions via straight-line ODE paths and iterative reflow. It supports generative image modeling (e.g., CIFAR-10) and unsupervised domain transfer, with Colab tutorials included.

Use cases

• train a rectified flow generative model on image datasets
• generate images in one step with a reflowed ODE model
• learn transport maps between two data distributions
• perform unsupervised image-to-image domain transfer
• reproduce results from the Rectified Flow ICLR 2023 paper
• learn flow-based generative models with straight trajectories

When to choose

• you want the reference implementation of the Rectified Flow paper
• you need fast one-step generation with better FID than fast diffusion models and more diversity than GANs
• you want to experiment with flow matching or reflow techniques
• you need ODE-based transport between distributions for domain transfer

When to avoid

• you need a production-ready, well-maintained library with a license and long-term support
• you want a general-purpose diffusion or flow-matching framework with broad model support
• you need Stable Diffusion one-step generation specifically (use the related InstaFlow repo instead)
• you require Windows or non-CUDA environments out of the box

Facets

library · maturity maintenance

machine-learning deep-learning image-processing machine-learning deep-learning image-processing artificial-intelligence python rectified-flow generative-modeling diffusion-models ode research-code image-generation domain-transfer iclr2023 linux gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="gnobitab/RectifiedFlow")