buoyancy99/diffusion-forcing
A PyTorch implementation of Diffusion Forcing, a generative training technique that combines autoregressive and diffusion paradigms for sequential data like video.
Not currently ranked — collecting fresh signals.
star history
This repository provides the official PyTorch implementation of Diffusion Forcing, a method that bridges next-token prediction (autoregressive) and full-sequence diffusion models. The technique trains a model to denoise subsets of tokens at varying noise levels simultaneously, enabling flexible generation across different timesteps. The codebase includes temporal attention mechanisms and supports applications in video generation, robotics, and planning domains.
Frequently asked
- What is buoyancy99/diffusion-forcing?
- A PyTorch implementation of Diffusion Forcing, a generative training technique that combines autoregressive and diffusion paradigms for sequential data like video.
- Is diffusion-forcing open source?
- Yes — buoyancy99/diffusion-forcing is an open-source project tracked on heatdrop.
- What language is diffusion-forcing written in?
- buoyancy99/diffusion-forcing is primarily written in Python.
- How popular is diffusion-forcing?
- buoyancy99/diffusion-forcing has 1.3k stars on GitHub.
- Where can I find diffusion-forcing?
- buoyancy99/diffusion-forcing is on GitHub at https://github.com/buoyancy99/diffusion-forcing.