rectified-flow-pytorch and flow-matching
These two tools are **competitors**: both provide PyTorch implementations of different but related approaches to continuous normalizing flows (rectified flows vs. flow matching) for generative modeling.
About rectified-flow-pytorch
lucidrains/rectified-flow-pytorch
Implementation of rectified flow and some of its followup research / improvements in Pytorch
This tool helps researchers and practitioners in machine learning to train and use Rectified Flow models, which are advanced generative models. You provide a collection of existing images, and the system learns to generate new, similar images. It's designed for machine learning researchers, deep learning engineers, and data scientists working on generative AI.
About flow-matching
keishihara/flow-matching
Flow Matching implemented in PyTorch
This tool helps machine learning engineers generate high-quality synthetic data, such as images or complex 2D data distributions. You provide a target data distribution, and it generates new samples that closely match its characteristics. This is ideal for researchers and practitioners in generative modeling who need to create realistic data for training or analysis.
Scores updated daily from GitHub, PyPI, and npm data. How scores work