DFLASH

// DFlash: Block Diffusion for Flash Speculative Decoding

dflash

DFlash: Block Diffusion for Flash Speculative Decoding

13EmergingUnknown

See Alternatives Compare...

License

MIT

Updated

Today

What it does

Paper | Blog | Models DFlash is a lightweight block diffusion model designed for speculative decoding. It enables efficient and high-quality parallel drafting. https://github.com/user-attachments/assets/5b29cabb-eb95-44c9-8ffe-367c0758de8c - openai/gpt-oss-20b: https://huggingface.co/z-lab/gpt-oss-20b-DFlash - Qwen3-4B: https://huggingface.co/z-lab/Qwen3-4B-DFlash-b16 - Qwen3-8B:

Getting Started

git

git clone https://github.com/z-lab/dflash

Links

Platforms

🪟windows🍎mac🐧linux

Install Difficulty

moderate

Built With

python

Community Reactions

Similar Tools

See all alternatives →

36Emerging

Openvpn_webpanel_manager

A powerful, self-hosted web panel for managing OpenVPN servers, users, resellers (sub-admins), and multi-node deploym...

Active190open-source

36Emerging

Hink

Link Shortener for Hackers

Active158open-source

36Emerging

HackerBook

Hacker Book - COMMUNITY, ALL THE HN ARE BELONG TO YOU. An unkillable, static offline archive of all of Hacker News.

Active187open-source

36Emerging

Obsidian Life Tracker Base View

Capture and visualize the data that matters in your life

Active179open-source

36Emerging

Vyuh_node_flow

A flexible, high-performance node-based flow editor for Flutter. Build visual programming interfaces, workflow editor...

Active171open-source

36Emerging

Db Studio

The modern pgAdmin alternative that works with every database.

Active171open-source