{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2025,10,8]],"date-time":"2025-10-08T05:41:19Z","timestamp":1759902079965,"version":"build-2065373602"},"reference-count":54,"publisher":"IEEE","license":[{"start":{"date-parts":[[2025,9,2]],"date-time":"2025-09-02T00:00:00Z","timestamp":1756771200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2025,9,2]],"date-time":"2025-09-02T00:00:00Z","timestamp":1756771200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2025,9,2]]},"DOI":"10.1109\/cluster59342.2025.11186491","type":"proceedings-article","created":{"date-parts":[[2025,10,7]],"date-time":"2025-10-07T17:35:09Z","timestamp":1759858509000},"page":"1-11","source":"Crossref","is-referenced-by-count":0,"title":["SplitQuant: Resource-Efficient LLM Offline Serving on Heterogeneous GPUs via Phase-Aware Model Partition and Adaptive Quantization"],"prefix":"10.1109","author":[{"given":"Juntao","family":"Zhao","sequence":"first","affiliation":[{"name":"The University of Hong Kong,Hong Kong"}]},{"given":"Borui","family":"Wan","sequence":"additional","affiliation":[{"name":"The University of Hong Kong,Hong Kong"}]},{"given":"Yanghua","family":"Peng","sequence":"additional","affiliation":[{"name":"ByteDance Inc.,USA"}]},{"given":"Haibin","family":"Lin","sequence":"additional","affiliation":[{"name":"ByteDance Inc.,USA"}]},{"given":"Chuan","family":"Wu","sequence":"additional","affiliation":[{"name":"The University of Hong Kong,Hong Kong"}]}],"member":"263","reference":[{"journal-title":"OpenAI","article-title":"Chatgpt: Optimizing language models for dialogue","year":"2023","key":"ref1"},{"key":"ref2","article-title":"Bloom: A 176b-parameter open-access multilingual language model","volume":"abs\/2211.05100","author":"Scao","year":"2022","journal-title":"ArXiv"},{"key":"ref3","article-title":"Opt: Open pre-trained transformer language models","volume":"abs\/2205.01068","author":"Zhang","year":"2022","journal-title":"ArXiv"},{"key":"ref4","doi-asserted-by":"publisher","DOI":"10.5040\/9781501365072.09396"},{"key":"ref5","doi-asserted-by":"publisher","DOI":"10.1145\/3600006.3613165"},{"key":"ref6","article-title":"SGLang: Efficient execution of structured language model programs","volume-title":"The Thirty-eighth Annual Conference on Neural Information Processing Systems","author":"Zheng","year":"2024"},{"volume-title":"TensorRT-LLM Backend for Triton Inference Server.","year":"2023","author":"Corporation","key":"ref7"},{"journal-title":"Megatron-lm: Training multi-billion parameter language models using model parallelism","year":"2020","author":"Shoeybi","key":"ref8"},{"key":"ref9","doi-asserted-by":"publisher","DOI":"10.48550\/arxiv.1811.06965"},{"key":"ref10","article-title":"Awq: Activation-aware weight quantization for 11 m compression and acceleration","volume":"abs\/2306.00978","author":"Lin","year":"2023","journal-title":"ArXiv"},{"key":"ref11","article-title":"Gptq: Accurate post-training quantization for generative pre-trained transformers","volume":"abs\/2210.17323","author":"Frantar","year":"2022","journal-title":"ArXiv"},{"key":"ref12","article-title":"Pipeline parallelism for inference on heterogeneous edge computing","volume":"abs\/2110.14895","author":"Hu","year":"2021","journal-title":"ArXiv"},{"key":"ref13","article-title":"Smoothquant: Accurate and efficient post-training quantization for large language models","volume-title":"International Conference on Machine Learning","author":"Xiao","year":"2023"},{"journal-title":"Deliver high performance ml inference with aws inferentia","year":"2019","author":"Hutt","key":"ref14"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2010.11929"},{"key":"ref16","article-title":"Bert: Pre-training of deep bidirectional transformers for language understanding","author":"Devlin","year":"2018","journal-title":"arXiv preprint"},{"key":"ref17","article-title":"Flexgen: High-throughput generative inference of large language models with a single gpu","volume-title":"Proceedings of the 40th International Conference on Machine Learning","author":"Sheng","year":"2023"},{"journal-title":"RyokoAI","article-title":"Sharegpt52k","year":"2021","key":"ref18"},{"key":"ref19","doi-asserted-by":"publisher","DOI":"10.1109\/SC41404.2022.00051"},{"key":"ref20","article-title":"Orca: A distributed serving system for Transformer-Based generative models","volume-title":"16th USENIX Symposium on Operating Systems Design and Implementation (OSDI 22)","author":"Yu","year":"2022"},{"journal-title":"Sarathi: Efficient 11 m inference by piggybacking decodes with chunked prefills","year":"2023","author":"Agrawal","key":"ref21"},{"key":"ref22","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/P17-1099"},{"journal-title":"The llama 3 herd of models","year":"2024","author":"Grattafiori","key":"ref23"},{"key":"ref24","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.acl-long.859"},{"key":"ref25","article-title":"Zeroquant: Efficient and affordable post-training quantization for largescale transformers","author":"Yao","year":"2022","journal-title":"in Advances in Neural Information Processing Systems"},{"key":"ref26","article-title":"Llm. int8 (): 8-bit matrix multiplication for transformers at scale","author":"Dettmers","year":"2022","journal-title":"arXiv preprint"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.1121\/1.2016299"},{"journal-title":"Pointer sentinel mixture models","year":"2016","author":"Merity","key":"ref28"},{"key":"ref29","doi-asserted-by":"publisher","DOI":"10.21236\/ADA273556"},{"key":"ref30","article-title":"Exploring the limits of transfer learning with a unified text-to-text transformer","author":"Raffel","year":"2019","journal-title":"arXiv e-prints"},{"key":"ref31","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/P16-1144"},{"journal-title":"Think you have solved question answering? try arc, the ai2 reasoning challenge","year":"2018","author":"Clark","key":"ref32"},{"key":"ref33","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v34i05.6239"},{"key":"ref34","doi-asserted-by":"publisher","DOI":"10.1145\/3710848.3710871"},{"journal-title":"Qserve: W4a8kv4 quantization and system co-design for efficient 1 lm serving","year":"2024","author":"Lin","key":"ref35"},{"key":"ref36","article-title":"Spqr: A sparsequantized representation for near-lossless 11 m weight compression","volume":"abs\/2306.03078","author":"Dettmers","year":"2023","journal-title":"ArXiv"},{"key":"ref37","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2019.00038"},{"key":"ref38","article-title":"Adaptive message quantization and parallelization for distributed full-graph gnn training","volume":"abs\/2306.01381","author":"Wan","year":"2023","journal-title":"ArXiv"},{"key":"ref39","article-title":"Alpa: Automating inter-and \\{IntraOperator\\} parallelism for distributed deep learning","volume-title":"16th USENIX Symposium on Operating Systems Design and Implementation (OSDI 22)","author":"Zheng","year":"2022"},{"journal-title":"Gurobi Optimization, LLC","article-title":"Gurobi Optimizer Reference Manual","year":"2023","key":"ref40"},{"article-title":"Gptqmodel.","volume-title":"ModelCloud.ai and","year":"2024","key":"ref41"},{"article-title":"Llm compressor: Efficient compression for large language models.","volume-title":"vLLM Team","year":"2025","key":"ref42"},{"journal-title":"Flashattention: Fast and memory-efficient exact attention with io-awareness","year":"2022","author":"Dao","key":"ref43"},{"key":"ref44","article-title":"Pytorch: An imperative style, high-performance deep learning library","author":"Paszke","year":"2019","journal-title":"in Neural Information Processing Systems"},{"journal-title":"Qwen2.5 technical report","year":"2025","author":"Yang","key":"ref45"},{"journal-title":"Hexgen: Generative inference of large language model over heterogeneous environment","year":"2024","author":"Jiang","key":"ref46"},{"key":"ref47","first-page":"1693","article-title":"Teaching machines to read and comprehend","author":"Hermann","year":"2015","journal-title":"in NIPS"},{"key":"ref48","doi-asserted-by":"publisher","DOI":"10.1145\/3694715.3695964"},{"journal-title":"Hetegen: Heterogeneous parallel inference for large language models on resourceconstrained devices","year":"2024","author":"Zhao","key":"ref49"},{"journal-title":"Dovetail: A cpu\/gpu heterogeneous speculative decoding for llm inference","year":"2024","author":"Zhang","key":"ref50"},{"key":"ref51","doi-asserted-by":"publisher","DOI":"10.1145\/3669940.3707215"},{"journal-title":"Hexgen-2: Disaggregated generative inference of 11 ms in heterogeneous environment","year":"2025","author":"Jiang","key":"ref52"},{"key":"ref53","article-title":"Atom: Low-bit quantization for efficient and accurate llm serving","volume-title":"Proceedings of Machine Learning and Systems","author":"Zhao","year":"2024"},{"journal-title":"Qqq: Quality quattuor-bit quantization for large language models","year":"2024","author":"Zhang","key":"ref54"}],"event":{"name":"2025 IEEE International Conference on Cluster Computing (CLUSTER)","start":{"date-parts":[[2025,9,2]]},"location":"United Kingdom","end":{"date-parts":[[2025,9,5]]}},"container-title":["2025 IEEE International Conference on Cluster Computing (CLUSTER)"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/11186399\/11186452\/11186491.pdf?arnumber=11186491","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,10,8]],"date-time":"2025-10-08T05:02:39Z","timestamp":1759899759000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11186491\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,9,2]]},"references-count":54,"URL":"https:\/\/doi.org\/10.1109\/cluster59342.2025.11186491","relation":{},"subject":[],"published":{"date-parts":[[2025,9,2]]}}}