lebellig/discrete-fm
Educational implementation of the Discrete Flow Matching paper
Implements discrete flow matching for image generation by learning conditional velocity fields over discrete state transitions, with formulas directly traceable to the original paper. The notebook-based approach demonstrates the core algorithm on image datasets, bridging continuous flow matching theory to discrete generative tasks. Integrates with PyTorch and Jupyter for interactive exploration of diffusion-style generation on categorical data.
133 stars. No commits in the last 6 months.
Stars
133
Forks
7
Language
Jupyter Notebook
License
—
Category
Last pushed
Aug 26, 2024
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/generative-ai/lebellig/discrete-fm"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
AlejandroMllo/action_flow_matching
Code for the paper "Action Flow Matching for Continual Robot Learning" presented at Robotics:...
IRedDragonICY/auraflow_rrf
Relativistic Resonant Flow (RRF) Sampler: A novel discretization method for Rectified Flow...
lebellig/flow-matching
Annotated Flow Matching paper
AngeClementAkazan/Sequential-FeatureForestFlow
This repository contains the code source of the Heterogeneous Sequential Feature Forest Flow...
liyuantsao/BFSR
The official repository of BFSR: "Boosting Flow-based Generative Super-Resolution Models via...