Jiamin Li

Jiamin Li, Qiang Su, Yitao Yang et al.

Large language models, such as OpenAI's ChatGPT, have demonstrated exceptional language understanding capabilities in various NLP tasks. Sparsely activated mixture-of-experts (MoE) has emerged as a promising solution for scaling models while maintaining a constant number of computational operations. Existing MoE model adopts a fixed gating network where each token is computed by the same number of experts. However, this approach contradicts our intuition that the tokens in each sequence vary in terms of their linguistic complexity and, consequently, require different computational costs. Little is discussed in prior research on the trade-off between computation per token and model performance. This paper introduces adaptive gating in MoE, a flexible training strategy that allows tokens to be processed by a variable number of experts based on expert probability distribution. The proposed framework preserves sparsity while improving training efficiency. Additionally, curriculum learning is leveraged to further reduce training time. Extensive experiments on diverse NLP tasks show that adaptive gating reduces at most 22.5% training time while maintaining inference quality. Moreover, we conduct a comprehensive analysis of the routing decisions and present our insights when adaptive gating is used.

Jiamin Li, Hong Xu, Yibo Zhu et al.

Companies build separate training and inference GPU clusters for deep learning, and use separate schedulers to manage them. This leads to problems for both training and inference: inference clusters have low GPU utilization when the traffic load is low; training jobs often experience long queueing time due to lack of resources. We introduce Aryl, a new cluster scheduler to address these problems. Aryl introduces capacity loaning to loan idle inference GPU servers for training jobs. It further exploits elastic scaling that scales a training job's GPU allocation to better utilize loaned resources. Capacity loaning and elastic scaling create new challenges to cluster management. When the loaned servers need to be returned, we need to minimize the number of job preemptions; when more GPUs become available, we need to allocate them to elastic jobs and minimize the job completion time (JCT). Aryl addresses these combinatorial problems using principled heuristics. It introduces the notion of server preemption cost which it greedily reduces during server reclaiming. It further relies on the JCT reduction value defined for each additional worker for an elastic job to solve the scheduling problem as a multiple-choice knapsack problem. Prototype implementation on a 64-GPU testbed and large-scale simulation with 15-day traces of over 50,000 production jobs show that Aryl brings 1.53x and 1.50x reductions in average queuing time and JCT, and improves cluster usage by up to 26.9% over the cluster scheduler without capacity loaning or elastic scaling.

Haonan Yan, Xiaoguang Li, Hui Li et al.

Public intelligent services enabled by machine learning algorithms are vulnerable to model extraction attacks that can steal confidential information of the learning models through public queries. Though there are some protection options such as differential privacy (DP) and monitoring, which are considered promising techniques to mitigate this attack, we still find that the vulnerability persists. In this paper, we propose an adaptive query-flooding parameter duplication (QPD) attack. The adversary can infer the model information with black-box access and no prior knowledge of any model parameters or training data via QPD. We also develop a defense strategy using DP called monitoring-based DP (MDP) against this new attack. In MDP, we first propose a novel real-time model extraction status assessment scheme called Monitor to evaluate the situation of the model. Then, we design a method to guide the differential privacy budget allocation called APBA adaptively. Finally, all DP-based defenses with MDP could dynamically adjust the amount of noise added in the model response according to the result from Monitor and effectively defends the QPD attack. Furthermore, we thoroughly evaluate and compare the QPD attack and MDP defense performance on real-world models with DP and monitoring protection.