2025

1. VLDB

   mLoRA: Fine-Tuning LoRA Adapters via Highly-Efficient Pipeline Parallelism in Multiple GPUs

   Zhengmao Ye, Dengchun Li, Zetao Hu, and 7 more authors

   2025

2024

1. ICML

   BadPart: Unified Black-box Adversarial Patch Attacks against Pixel-wise Regression Tasks

   Zhiyuan Cheng, Zhaoyi Liu, Tengda Guo, and 4 more authors

   In Forty-first International Conference on Machine Learning, 2024
2. VLDB

   GPTuner: A Manual-Reading Database Tuning System via GPT-Guided Bayesian Optimization

   Jiale Lao, Yibo Wang, Yufei Li, and 6 more authors

   Proc. VLDB Endow., May 2024

   Abs

   Modern database management systems (DBMS) expose hundreds of configurable knobs to control system behaviours. Determining the appropriate values for these knobs to improve DBMS performance is a long-standing problem in the database community. As there is an increasing number of knobs to tune and each knob could be in continuous or categorical values, manual tuning becomes impractical. Recently, automatic tuning systems using machine learning methods have shown great potentials. However, existing approaches still incur significant tuning costs or only yield sub-optimal performance. This is because they either ignore the extensive domain knowledge available (e.g., DBMS manuals and forum discussions) and only rely on the runtime feedback of benchmark evaluations to guide the optimization, or they utilize the domain knowledge in a limited way. Hence, we propose GPTuner, a manual-reading database tuning system that leverages domain knowledge extensively and automatically to optimize search space and enhance the runtime feedback-based optimization process. Firstly, we develop a Large Language Model (LLM)-based pipeline to collect and refine heterogeneous knowledge, and propose a prompt ensemble algorithm to unify a structured view of the refined knowledge. Secondly, using the structured knowledge, we (1) design a workload-aware and training-free knob selection strategy, (2) develop a search space optimization technique considering the value range of each knob, and (3) propose a Coarse-to-Fine Bayesian Optimization Framework to explore the optimized space. Finally, we evaluate GPTuner under different benchmarks (TPC-C and TPC-H), metrics (throughput and latency) as well as DBMS (PostgreSQL and MySQL). Compared to the state-of-the-art approaches, GPTuner identifies better configurations in 16x less time on average. Moreover, GPTuner achieves up to 30% performance improvement (higher throughput or lower latency) over the best-performing alternative.

3. VLDB

   DLRover-RM: Resource Optimization for Recommendation Models Training in Cloud

   Qinlong Wang*, Tingfeng Lan*, Yinghao Tang, and 9 more authors

   May 2024
4. SIGMOD

   A Demonstration of GPTuner: A GPT-Based Manual-Reading Database Tuning System

   Jiale Lao, Yibo Wang, Yufei Li, and 7 more authors

   In Companion of the 2024 International Conference on Management of Data, May 2024

   Abs

   Selecting appropriate values for the configurable knobs of Database Management Systems (DBMS) is crucial to improve performance. But because such complexity has surpassed the abilities of even the best human experts, database community turns to machine learning (ML)-based automatic tuning systems. However, these systems still incur significant tuning costs or only yield sub-optimal performance, attributable to their overly high reliance on black-box optimization and an oversight of domain knowledge. This paper demonstrates GPTuner, a manual-reading database tuning system that leverages Large Language Model (LLM) to bridge the gap between black-box optimization and white-box domain knowledge. This demonstration empowers (1) regular users with limited tuning experience to gain qualitative insights on the features of knobs, and optimize their DBMS performance automatically and efficiently, (2) database administrators and experts to further enhance GPTuner by simply contributing their invaluable tuning suggestions in natural language. Finally, we offer visitors the opportunity to explore a range of DBMS and optimization metrics, coupled with the flexibility to tailor their target workloads to their specific needs. Our full implementation and a video demonstration are available at https://github.com/SolidLao/GPTuner and https://youtu.be/Hz5Zck-9TlA, respectively.