Xupeng Miao

Email: xupeng@purdue.edu

photo_me.jpg

Xupeng Miao is a Kevin C. and Suzanne L. Kahn New Frontiers Assistant Professor in the Department of Computer Science at Purdue University. Before that, he was a Post Doctoral Fellow working with Prof. Zhihao Jia and Prof. Tianqi Chen in Catalyst Group and Parallel Data Lab at Computer Science Department of Carnegie Mellon University. He received his Ph.D. degree in computer science from Peking University in June 2022, supervised by Prof. Bin Cui, and his Bachelor’s degree from Northeastern University. He is the creator of Hetu, a highly efficient distributed deep learning system, and continuously leading the team development. He is broadly interested in machine learning systems, data management and distributed computing.

Prospective students: I am actively looking for strong and self-motivated PhD/MS students, postdocs, and (remote) interns interested in building systems for machine learning to join my group. If you are interested in working with me, please read the FAQ info and send me an email with your CV and transcripts.

News

Oct 2, 2024 Helix and GraphPipe were accepted by ASPLOS 2025. :tada:
Aug 15, 2024 Our paper on memory-efficient PEFT won the Outstanding Paper Award of ACL 2024! :trophy:
Aug 13, 2024 I am honored to serve as the Artificat Evaluation Co-Chair of MLSys 2025! :mega:
Aug 6, 2024 One paper on distributed LLM training was accepted by SOSP 2024. :tada:
Jun 18, 2024 I was awared WAIC 2024 Yunfan Award · Bright Stars! :confetti_ball:
Apr 24, 2024 We will lanuch a tutorial on efficient LLM serving in ICML 2024. :mega:
Apr 23, 2024 I was invited to give a talk at the ASPLOS’24 XTensor workshop. :mega:
Apr 4, 2024 I was invited to give a talk at the MLSys’24 Young Professionals Symposium. :mega:
Feb 29, 2024 SpecInfer was accepted by ASPLOS 2024. :tada:
Feb 28, 2024 SpotServe has been selected for the Distinguished Artifact Award at ASPLOS 2024! :trophy:
Feb 2, 2024 We will lanuch a tutorial on data managment for LLM in SIGMOD 2024. :mega:
Dec 23, 2023 We announce a survey about efficient generative LLM serving on arXiv. :mega:
Dec 7, 2023 One paper on distributed training over spot instances was accepted by NSDI 2024. :tada:
Nov 7, 2023 One paper about LLM serving over preemptive instances was accepted by ASPLOS 2024. :tada:
May 16, 2023 We announce the first speculative LLM inference engine called SpecInfer. :mega:
May 13, 2023 Three papers were accepted by VLDB 2023. :tada:
Mar 23, 2023 One paper was accepted by OSDI 2023. :tada:

Selected Publications

  1. ASPLOS
    SpotServe: Serving Generative Large Language Models on Preemptible Instances (Distinguished Artifact Award)
    Xupeng Miao, Chunan Shi, Jiangfei Duan,  Xiaoli Xi and 3 more authors
    Proceedings of ASPLOS Conference 2024
  2. ASPLOS
    SpecInfer: Accelerating Generative Large Language Model Serving with Speculative Inference and Token Tree Verification
    Xupeng Miao, Gabriele Oliaro, Zhihao Zhang,  Xinhao Cheng and 10 more authors
    Proceedings of ASPLOS Conference 2024
  3. NSDI
    Parcae: Proactive, Liveput-Optimized DNN Training on Preemptible Instances
    Jiangfei Duan1, Ziang Song1Xupeng Miao1,  Xiaoli Xi and 4 more authors
    Proceedings of NSDI Conference 2024
  4. VLDB
    SDPipe: A Semi-Decentralized Framework for Heterogeneity-aware Pipeline-parallel Training
    Xupeng Miao, Yining Shi, Zhi Yang,  Bin Cui and 1 more author
    Proc. VLDB Endow. 2023
  5. VLDB
    Galvatron: Efficient Transformer Training over Multiple GPUs Using Automatic Parallelism
    Xupeng Miao, Yujie Wang, Youhe Jiang,  Chunan Shi and 3 more authors
    Proc. VLDB Endow. 2023
  6. VLDB
    HET: Scaling out Huge Embedding Model Training via Cache-enabled Distributed Framework (Best Scalable Data Science Paper Award)
    Xupeng Miao, Hailin Zhang, Yining Shi,  Xiaonan Nie and 3 more authors
    Proc. VLDB Endow. 2022
  7. SIGMOD
    HET-GMP: A Graph-based System Approach to Scaling Large Embedding Model Training
    Xupeng Miao, Yining Shi, Hailin Zhang,  Xin Zhang and 3 more authors
    In Proceedings of SIGMOD Conference 2022
  8. SIGMOD
    Heterogeneity-Aware Distributed Machine Learning Training via Partial Reduce
    Xupeng Miao, Xiaonan Nie, Yingxia Shao,  Zhi Yang and 3 more authors
    In Proceedings of SIGMOD Conference 2021

Teaching