Metraipolis › archive › dc_46b301fbec6d

MG-Cache: Multi-Granularity Caching for Accelerating Diffusion Transformer Inference — arXiv:2611.11365

Source
arXiv (en)
Original
https://arxiv.org/abs/2611.11365
License (as archived)
CC0-1.0
Archived
2026-11-10T18:46:39.675Z
SHA-256 of the archived text
d85f8e0203b3ce680461b2c7abb5b0754606ce156ebf8888832c94a54a79bb14
Stable URL of this page
https://metraipolis.com/documento/dc_46b301fbec6d

How to cite

MG-Cache: Multi-Granularity Caching for Accelerating Diffusion Transformer Inference — arXiv:2611.11365 — arXiv (en), https://arxiv.org/abs/2611.11365 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_46b301fbec6d (Metraipolis archive dc_46b301fbec6d, sha256 d85f8e0203b3ce680461b2c7abb5b0754606ce156ebf8888832c94a54a79bb14)
BibTeX
@misc{dc46b301fbec6d,
  title = {MG-Cache: Multi-Granularity Caching for Accelerating Diffusion Transformer Inference — arXiv:2611.11365},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2611.11365},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_46b301fbec6d; sha256 d85f8e0203b3ce680461b2c7abb5b0754606ce156ebf8888832c94a54a79bb14},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.11365. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Diffusion Transformer (DiT) models, while having become a powerful backbone for generative synthesis, suffer from high inference cost on generating high-resolution images. Diffusion caching has emerged as a promising acceleration technique to reduce inference cost. However, existing caching strategies fail to accurately identify the varying importance of different computational modules and rarely coordinate cache reusing across the timestep and layer granularities, making it difficult to control error accumulation. To address these limitations, we propose multi-granularity caching (MG-Cache) for DiTs, a novel and fully training-free approach to exploit computational redundancy both on timesteps and Transformer layers. Our core innovation contains a two-stage caching strategy: an offline timestep profiling analysis that identifies redundant timesteps for cache reuse, followed by an adaptive layer-wise caching strategy that combines input stability with importance of each Transformer layer to selectively cache stable Attention and MLP submodules. Extensive experiments on mainstream open-source DiT models validate the effectiveness of our approach, achieving significant acceleration while maintaining high fidelity to the uncached output, with up to a 4.18x acceleration on HunyuanVideo.

Tema degli esploratori: calcolo e infrastrutture (elenco arXiv cs.PF, nuovi lavori del giorno).
Categorie arXiv: Machine Learning (cs.LG) ; Performance (cs.PF).
Pagina del lavoro: https://arxiv.org/abs/2611.11365 — 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 2026-11-10T18:38:37.395Z dalla pagina https://arxiv.org/list/cs.PF/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).