Abstract:LLMs often struggle with memory-constrained deployment on consumer-grade hardware due to their massive parameter sizes. While existing solutions such as model compression and offloading improve deployment feasibility, they often suffer from substantial accuracy degradation or severe throughput bottlenecks. Recent error compensation methods recover accuracy through auxiliary LoRA-style branches, and we observe that these branches are inherently amenable to offloading: they require substantial parameter storage but access only a small subset of compensation parameters during each inference step. Motivated by this opportunity, we propose HCInfer, a heterogeneous inference system that offloads residual compensation to the CPU while executing the compressed backbone on the GPU, and further introduces an asynchronous compensation pipeline and sensitivity-aware dynamic rank allocation to hide compensation overhead and maximize accuracy recovery. Experimental results show that HCInfer achieves a maximum accuracy improvement of 5.2% on downstream tasks compared to compression model and sustaining a maximum speedup of 10.4x compared to full-precision model.
| Subjects: | Machine Learning (cs.LG) |
| Cite as: | arXiv:2605.05819 [cs.LG] |
| (or arXiv:2605.05819v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2605.05819 arXiv-issued DOI via DataCite (pending registration) |
From: Xiangwen Zhuge [view email]
[v1]
Thu, 7 May 2026 07:57:23 UTC (601 KB)
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。