Metraipolis › archive › dc_98f73bdda0ca
TRACE: Transformer Residual Activation Codebook Encoding for Communication-efficient Distributed Edge Inference — arXiv:2610.22435
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.22435
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-19T08:42:47.615Z
- SHA-256 of the archived text
f6f53ca63ad7eb839a218cccc9aa7386748cfaa64ca1ef8707450812fc7c760c- Stable URL of this page
https://metraipolis.com/documento/dc_98f73bdda0ca
How to cite
TRACE: Transformer Residual Activation Codebook Encoding for Communication-efficient Distributed Edge Inference — arXiv:2610.22435 — arXiv (en), https://arxiv.org/abs/2610.22435 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_98f73bdda0ca (Metraipolis archive dc_98f73bdda0ca, sha256 f6f53ca63ad7eb839a218cccc9aa7386748cfaa64ca1ef8707450812fc7c760c)
BibTeX
@misc{dc98f73bdda0ca,
title = {TRACE: Transformer Residual Activation Codebook Encoding for Communication-efficient Distributed Edge Inference — arXiv:2610.22435},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.22435},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_98f73bdda0ca; sha256 f6f53ca63ad7eb839a218cccc9aa7386748cfaa64ca1ef8707450812fc7c760c},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.22435. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Distributed Transformer inference across co-located edge devices is fundamentally communication-bound, as exchanging activations between sequence-parallel partitions dominates inference latency under commodity edge bandwidths. Existing compression schemes reduce this cost but retain a structural bottleneck: the received data must still undergo key/value projection before attention can proceed, serializing computation behind communication. This paper presents Transformer Residual Activation Codebook Encoding (Trace), a framework that addresses these limitations. Trace places a Vector Quantizer (VQ) at the post-attention residual of every Transformer layer and precomputes codebook-indexed key/value representations. During inference, exchanged activations are encoded into compact codebook indices, allowing receivers to directly recover keys and values through table lookup instead of executing the downstream computations required to derive them. By eliminating the received data downstream computation, Trace enables communication--computation overlap and reduces a key latency bottleneck in distributed Transformer inference. A hierarchical clustering further structures the learned codebook into a multi-resolution hierarchy, enabling runtime adaptation of the communication bitrate without retraining. The framework supports two operating variants: one prioritizing accuracy and the other balancing accuracy and efficiency through buffer-compatible quantization. Experiments on ViT-Base (ImageNet-1K and CIFAR-100) and ViViT-B (UCF-101) show that Trace achieves up to 1.89% higher accuracy and up to 1.27x end-to-end inference speedup compared with the state-of-the-art VQ-based distributed inference framework. Compared with uncompressed sequence-parallel inference, Trace reduces communication overhead by up to 22,624x. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno). Categorie arXiv: Machine Learning (cs.LG) ; Artificial Intelligence (cs.AI); Distributed, Parallel, and Cluster Computing (cs.DC). Pagina del lavoro: https://arxiv.org/abs/2610.22435 — 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 2026-10-19T08:39:10.104Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).