nunchaku-ai/nunchaku

[ICLR2025 Spotlight] SVDQuant: Absorbing Outliers by Low-Rank Components for 4-Bit Diffusion Models

57
/ 100
Established

SVDQuant decomposes weight outliers into separate low-rank components during 4-bit quantization, enabling aggressive quantization without quality loss across diffusion and image generation models. The engine provides native integration with ComfyUI and Hugging Face, supports dynamic LoRA composition and asynchronous CPU offloading to reduce VRAM requirements, and targets both consumer GPUs (20-series+) and enterprise hardware through optimized kernels for INT4 and NVFP4 precisions.

3,724 stars.

No Package No Dependents
Maintenance 13 / 25
Adoption 10 / 25
Maturity 16 / 25
Community 18 / 25

How are scores calculated?

Stars

3,724

Forks

229

Language

Python

License

Apache-2.0

Last pushed

Mar 07, 2026

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/diffusion/nunchaku-ai/nunchaku"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.