nunchux-ai/nunchaku
View on GitHub ↗ ★ 3,963 Python
[ICLR2025 Spotlight] SVDQuant: Absorbing Outliers by Low-Rank Components for 4-Bit Diffusion Models
4 mentions on Hacker News by 3 people: 0 stories, 4 comments · first 2024-11-09 · latest 2025-04-30
20082026
-
Have you seen https://github.com/mit-han-lab/nunchaku for running flux on lower VRAM and thus faster performance? I have put up a PR https://github.com/ZiYang-xie/WorldGen/pull/7
comment by Maxious · 2025-04-30 · in WorldGen: Open-source 3D scene generator for Game/VR/XR -
It has already released as in https://github.com/mit-han-lab/nunchaku?tab=readme-ov-file#c... .
comment by lmxyy · 2025-02-23 · in SVDQuant+NVFP4: 4× Smaller, 3× Faster FLUX with 16-bit Quality on Blackwell GPUs -
…INT4. Try our interactive demo below or at https://svdquant.mit.edu/ ! Our code is all available at https://github.com/mit-han-lab/nunchaku !
comment by lmxyy · 2025-02-22 · in SVDQuant+NVFP4: 4× Smaller, 3× Faster FLUX with 16-bit Quality on Blackwell GPUs -
https://github.com/mit-han-lab/nunchaku
comment by superkuh · 2024-11-09 · in SVDQuant: 4-Bit Quantization Powers 12B Flux on a 16GB 4090 GPU with 3x Speedup