Logo image
PipeSD: An Efficient Cloud-Edge Collaborative Pipeline Inference Framework with Speculative Decoding
Conference proceeding   Open access

PipeSD: An Efficient Cloud-Edge Collaborative Pipeline Inference Framework with Speculative Decoding

Yunhe Pan, Yunqi Gao, Bing Hu, Mahdi Boloursaz Mashhadi, Yitong Duan, Pei Xiao and Yanfeng Zhang
Proceedings of the 43rd International Conference on Machine Learning, Seoul, South Korea. PMLR 306
The 43rd International Conference on Machine Learning (ICML 26) (Seoul, South Korea, 06/07/2026–11/07/2026)
30/04/2026

Abstract

Speculative decoding can significantly accelerate LLM inference, especially given that its cloud-edge collaborative deployment offers cloud work-load offloading, offline robustness, and privacy enhancement. However, existing collaborative inference frameworks with speculative decoding are constrained by (i) sequential token generation and communication with low resource utilization , and (ii) inflexible cloud non-autoregressive verification (NAV) triggering that induces premature verification or costly rollbacks. In this paper, we propose PipeSD, an efficient cloud-edge collaborative pipeline inference framework with speculative decoding. PipeSD overlaps token generation and communication by a token-batch pipeline scheduling mechanism optimized by dynamic programming, and improves verification flexibility through a dual-threshold NAV triggering mechanism with a lightweight Bayesian optimization autotuner. We implement PipeSD using llama-cpp-python, PyTorch, and FastAPI, and evaluate it on a real-world cloud-edge testbed with two draft-target model pairs across four scenarios. Results show that PipeSD consistently outperforms state-of-the-art baselines, achieving 1.16×–2.16× speedup and reducing energy consumption by 14.3%–25.3%. Our code is available at https://anonymous.4open. science/r/PipeSD.
pdf
PipeSD3.56 MBDownloadView
Author's Accepted Manuscript Open Access
url
https://icml.cc/View
Event Website Conference website

Metrics

1 Record Views

Details

Logo image

Usage Policy