lucidrains/rectified-flow-pytorch
PyTorch library implementing rectified flow for generative image synthesis via flow matching.
Not currently ranked — collecting fresh signals.
star history
Rectified flow is a generative modeling technique based on flow matching that interpolates between noise and data distributions. This library provides a PyTorch implementation including UNet-based models, reflow training procedures, and sampling utilities for generating images. It supports training with the accelerate library and includes implementations of followup research improvements.
Frequently asked
- What is lucidrains/rectified-flow-pytorch?
- PyTorch library implementing rectified flow for generative image synthesis via flow matching.
- Is rectified-flow-pytorch open source?
- Yes — lucidrains/rectified-flow-pytorch is open source, released under the MIT license.
- What language is rectified-flow-pytorch written in?
- lucidrains/rectified-flow-pytorch is primarily written in Python.
- How popular is rectified-flow-pytorch?
- lucidrains/rectified-flow-pytorch has 473 stars on GitHub.
- Where can I find rectified-flow-pytorch?
- lucidrains/rectified-flow-pytorch is on GitHub at https://github.com/lucidrains/rectified-flow-pytorch.