Metraipolis › archive › dc_23397353be18

Chameleon: Dynamic Format Adapter for Efficient Diffusion — arXiv:2609.33496

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.33496
License (as archived)
CC0-1.0
Archived
2026-10-07T21:01:30.223Z
SHA-256 of the archived text
cd3a02dbda87815ede094f2187d87f7c9ff6ed8b6a418ca8eeb566a620cdbe6c
Stable URL of this page
https://metraipolis.com/documento/dc_23397353be18

How to cite

Chameleon: Dynamic Format Adapter for Efficient Diffusion — arXiv:2609.33496 — arXiv (en), https://arxiv.org/abs/2609.33496 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_23397353be18 (Metraipolis archive dc_23397353be18, sha256 cd3a02dbda87815ede094f2187d87f7c9ff6ed8b6a418ca8eeb566a620cdbe6c)
BibTeX
@misc{dc23397353be18,
  title = {Chameleon: Dynamic Format Adapter for Efficient Diffusion — arXiv:2609.33496},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.33496},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_23397353be18; sha256 cd3a02dbda87815ede094f2187d87f7c9ff6ed8b6a418ca8eeb566a620cdbe6c},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.33496. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Post-training quantization (PTQ) is the standard way to run modern diffusion models on memory-constrained accelerators, yet every existing diffusion PTQ scheme fixes the $\mathit{number\ format}$ in advance and only tunes the scale, zero point, or per-layer bit-width. At a fixed bit-width the best format depends on the distribution being encoded, and that distribution differs across weight channels, across layers, and along the diffusion timestep, where activation distributions slide from heavy-tailed and noise-dominated to tightly clustered and structured. We propose Chameleon, a PTQ framework that holds the bit-width fixed and treats the format itself as a discrete variable, chosen per weight channel and per (layer, timestep bucket) activation tensor. Activation formats come from {INT8, FP8 E4M3, FP8 E5M2, MXFP8, MXINT8}, selected ahead of time from two cheap statistics (empirical kurtosis and the closed-form diffusion SNR) and stored in a lookup table; weight formats come from {INT8, MXINT8} at 8 bits or {INT4, NF4, FP4 E2M1, MXINT4, MXFP4} at 4 bits, selected offline by reconstruction error. An architectural fork adapts the same selection layer to multi-step UNets, single-step distilled models, and Diffusion Transformers. Across SDXL, SDXL-Turbo, and PixArt-$\alpha$ on COCO-2014, Chameleon achieves the best FID in all six backbone $\times$ bit-width settings, with CLIP within 0.24 of the FP16 reference and the best of all quantized methods at $W_{4}A_{8}$.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in cs.AI).
Categorie arXiv: Machine Learning (cs.LG) ; Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV).
Pagina del lavoro: https://arxiv.org/abs/2609.33496 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-07T20:29:02.677Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).