检索结果-内蒙古大学图书馆

DELTA: Dynamically Optimizing GPU Memory beyond Tensor Recomputation

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Tang, Yu Wang, Chenyu Zhang, Yufan Liu, Yuliang Zhang, Xingcheng Qiao, Linbo Lai, Zhiquan Li, Dongsheng Science and Technology on Parallel and Distributed Processing Laboratory United States SenseTime Research Shanghai Artificial Intelligence Laboratory China

Training activations of deep neural networks occupy plenty of GPU memory, especially for large-scale deep neuralnetworks. However, the further development of deep neural networks is hampered by the limited GPU memory resource. Therefore, the optimal utilization of GPU memory resources is highly demanded. Swapping and recomputation are commonly applied to make better use of GPU memory in deep learning. As an emerging domain, several dilemmas remain: 1) The efficiency of recomputation is limited and swapping between GPU and CPU costs severe time delay;2) There still lacks a dynamic runtime memory manager of tensor swapping and tensor recomputation nowadays;3) Manually decisions for activations of training deep neural network require professional priors and experience. To remedy the above issues, we propose a novel memory manager named DELTA (Dynamic tEnsor offLoading and recompuTAtion). To the best of our knowledge, we are the first to propose a reasonable dynamic runtime manager on the combination of tensor swapping and tensor recomputation without user oversight. In DELTA, we firstly propose a filter algorithm to select the optimal tensors to be released out of GPU memory and secondly present a director algorithm to select a proper action for each of these tensors. Furthermore, prefetching and overlapping are deliberately considered to overcome the time cost caused by swapping and recomputing tensors. Experimental results show that DELTA not only saves 40%-70% of GPU memory, surpassing the state-of-the-art method to a great extent, but also gets comparable convergence results as the baseline with acceptable time delay. Also, DELTA gains 2.04× maximum batchsize when training ResNet-50 and 2.25× when training ResNet-101 compared with the baseline. Besides, comparisons between the swapping cost and recomputation cost in our experiments demonstrate the importance of making a reasonable decision on tensor swapping and tensor recomputation, which refutes the arguments in

关键词： Tensors

A Float-Point Arithmetic Logic Unit for High Performance RISC-V Processor

学校读者我要写书评

暂无评论

A Float-Point Arithmetic Logic Unit for High Performance RIS...

Frontiers technology of Information and Computer (ICFTIC), IEEE International Conference on

作者： Chunyuan Wei Chuan Yuan Yu Qi Bangjian Xu Xin Niu College of Computer Science and Electronic Engineering Hunan University Changsha China College of Electrical and Information Engineering Hunan University Changsha China Science and Technology on Parallel and Distributed Laboratory College of Computer National University of Defense Technology Changsha China

ISBN: (数字)9798331541750

ISBN: (纸本)9798331541767

With the rapid advancement of artificial intelligence, chips have become increasingly important. The emerging RISC-V instruction set gradually provides powerful computing support for this field. In this context, along with the computing requirements of deep learning, this paper presents the design of a high-performance floating-point arithmetic logic unit (FALU) that facilitates calculations with double-precision, single-precision, half-precision, and Bfloat16 precision data. This design is based on a single-channel algorithm with merged rounding. It improves and implements a composite adder that combines high and low bits. It also proposes a tree-like floating-point comparator based on the Kogge-Stone parallel prefix network. To ensure that the FALU components meet performance requirements, we undergo functional verification in the Vivado simulation environment. Operating at 1.47GHz under the 28nm CMOS process, the components achieve the predetermined performance indicators.

关键词： Deep learning Context Accuracy Instruction sets CMOS process Logic Artificial intelligence Adders Floating-point arithmetic

GNNRL-Smoothing: A Prior-Free Reinforcement Learning Model for Mesh Smoothing

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Wang, Zhichao C. Chen, Xinhai C. Gong, Chunye Y. Yang, Bo Deng, Liang Sun, Yufei F. Pang, Yufei F. Liu, Jie National University of Defense Technology Science and Technology on Parallel and Distributed Processing Laboratory China National University of Defense Technology Laboratory of Digitizing Software for Frontier Equipment China China Aerodynamics Research and Development Center China Nankai University College of Software China

Mesh smoothing methods can enhance mesh quality by eliminating distorted elements, leading to improved convergence in simulations. To balance the efficiency and robustness of traditional mesh smoothing process, previous approaches have employed supervised learning and reinforcement learning to train intelligent smoothing models. However, these methods heavily rely on labeled dataset or prior knowledge to guide the models’ learning. Furthermore, their limited capacity to enhance mesh connectivity often restricts the effectiveness of smoothing. In this paper, we first systematically analyze the learning mechanisms of recent intelligent smoothing methods and propose a prior-free reinforcement learning model for intelligent mesh smoothing. Our proposed model integrates graph neural networks with reinforcement learning to implement an intelligent node smoothing agent and introduces, for the first time, a mesh connectivity improvement agent. We formalize mesh optimization as a Markov Decision Process and successfully train both agents using Twin Delayed Deep Deterministic Policy Gradient and Double Dueling Deep Q-Network in the absence of any prior data or knowledge. We verified the proposed model on both 2D and 3D meshes. Experimental results demonstrate that our model achieves feature-preserving smoothing on complex 3D surface meshes. It also achieves state-of-the-art results among intelligent smoothing methods on 2D meshes and is 7.16 times faster than traditional optimization-based smoothing methods. Moreover, the connectivity improvement agent can effectively enhance the quality distribution of the mesh. © 2024, CC BY.

关键词： Deep reinforcement learning

WBA: A Warping-based Approach to Generating Imperceptible Adversarial Examples

学校读者我要写书评

暂无评论

WBA: A Warping-based Approach to Generating Imperceptible Ad...

IEEE International Conference on Trust, Security and Privacy in Computing and Communications (TrustCom)

作者： Chengyao Hua Shigeng Zhang Weiping Wang Zhankai Li Jian Zhang School of Computer Science and Engineering Central South University Changsha China Science and Technology on Parallel and Distributed Processing Laboratory (PDL)

The human can easily recognize the incongruous parts of an image, for example, perturbations unrelated to the image itself, but are poor at spotting the small geometric transformations. However, in terms of the robustness of deep neural networks (DNNs), the ability to properly recognize objects with small geometric transformations is still a challenge. In this work, we investigate the problem from the perspective of adversarial attacks: does the performance of DNNs degrade even when small geometric transformations are applied to images? To this end, we propose a novel adversarial attack method, called WBA, a Warping-Based Adversarial attack method, which does not introduce information independent of the original images but manipulates the existing pixels of the images by elastic warping transformations to generate adversarial examples that are imperceptible to the human eye. At the same time, existing adversarial attacks typically generate adversarial examples by modifying pixels in the spatial domain of the image, the addition of such perturbations introduces extra information unrelated to the image itself and is easily detected by the naked eyes. We demonstrate the effectiveness of WBA by extensive experiments on commonly used datasets, including MNIST, CIFAR10, and ImageNet. The results show that WBA can quickly generate adversarial examples with the highest adversarial strength, consumes less time, and can be comparable to optimization-based adversarial attack methods in image perception evaluation metrics such as LPIPS, SSIM, and far more than gradient direction-based iterative methods.

关键词： Measurement Deep learning Privacy Image recognition Perturbation methods Neural networks Robustness

Proposing an intelligent mesh smoothing method with graph neural networks

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Wang, Zhichao Chen, Xinhai Yan, Junjun Liu, Jie Science and Technology on Parallel and Distributed Processing Laboratory National University of Defense Technology Changsha410073 China Laboratory of Digitizing Software for Frontier Equipment National University of Defense Technology National University of Defense Technology Changsha410073 China

In CFD, mesh smoothing methods are commonly utilized to refine the mesh quality to achieve high-precision numerical simulations. Specifically, optimization-based smoothing is used for high-quality mesh smoothing, but it incurs significant computational overhead. Pioneer works improve its smoothing efficiency by adopting supervised learning to learn smoothing methods from high-quality meshes. However, they pose difficulty in smoothing the mesh nodes with varying degrees and also need data augmentation to address the node input sequence problem. Additionally, the required labeled high-quality meshes further limit the applicability of the proposed method. In this paper, we present GMSNet, a lightweight neural network model for intelligent mesh smoothing. GMSNet adopts graph neural networks to extract features of the node's neighbors and output the optimal node position. During smoothing, we also introduce a fault-tolerance mechanism to prevent GMSNet from generating negative volume elements. With a lightweight model, GMSNet can effectively smoothing mesh nodes with varying degrees and remain unaffected by the order of input data. A novel loss function, MetricLoss, is also developed to eliminate the need for high-quality meshes, which provides a stable and rapid convergence during training. We compare GMSNet with commonly used mesh smoothing methods on two-dimensional triangle meshes. The experimental results show that GMSNet achieves outstanding mesh smoothing performances with 5% model parameters of the previous model, and attains 13.56 times faster than optimization-based smoothing. © 2023, CC BY.

关键词： Graph neural networks

A survey on deep learning approaches for data integration in autonomous driving system

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Zhu, Xi Wang, Likang Zhou, Caifa Cao, Xiya Gong, Yue Chen, Lei Riemann Laboratory Huawei Technologies 2012 Laboratories China Department of Computer Science and Engineering Hong Kong University of Science and Technology Hong Kong Parallel Distributed Computing Laboratory Huawei Technologies 2012 Laboratories China

The perception module of self-driving vehicles relies on a multi-sensor system to understand its environment. Recent advancements in deep learning have led to the rapid development of approaches that integrate multi-sensory measurements to enhance perception capabilities. This paper surveys the latest deep learning integration techniques applied to the perception module in autonomous driving systems, categorizing integration approaches based on "what, how, and when to integrate." A new taxonomy of integration is proposed, based on three dimensions: multi-view, multi-modality, and multi-frame. The integration operations and their pros and cons are summarized, providing new insights into the properties of an "ideal" data integration approach that can alleviate the limitations of existing methods. After reviewing hundreds of relevant papers, this survey concludes with a discussion of the key features of an optimal data integration approach. © 2023, CC BY-NC-SA.

关键词： Autonomous vehicles

F3A: Fairness-Aware AI-Workloads Allocation Considering Multidimensional User Demands in JointCloud

学校读者我要写书评

暂无评论

F3A: Fairness-Aware AI-Workloads Allocation Considering Mult...

IEEE International Conference on Joint Cloud Computing (JCC)

作者： Jiacheng Yang Guodong Yi Fei Gao Peichang Shi Huaimin Wang National Key Laboratory of Parallel and Distributed Processing College of Computer Science National University of Defense Technology Changsha China Xiangjiang Lab Changsha China School of Advanced Interdisciplinary Studies Hunan University Of Technology and Business Changsha China

ISBN: (数字)9798350387339

ISBN: (纸本)9798350387346

With the rapid growth of large language models, cloud computing has become an indispensable component of the AI industry. Cloud service providers(CSPs) are establishing AI data centers to service AI workloads. In the face of this surging need for AI computing power, building a connected computing environment across various clouds and forming a JointCloud presents an attractive solution. However, scheduling AI tasks across multiple AI data centers within a JointCloud environment presents a significant challenge: how to balance users’ demands while ensuring CSPs’ fairness in scheduling. Existing research primarily focuses on optimizing scheduling quality with limited consideration for fairness. Therefore, this paper proposes a Fairness-Aware AI-Workloads Allocation method (F3A), a fair cross-cloud allocation technique for AI tasks. F3A utilizes Point and Token to reflect both the resource status and historical task allocations of AI data centers, enabling the consideration of users’ multidimensional demands and facilitating fair task allocation across multiple centers. In order to better assess the fairness of scheduling, we also devised a fairness indicator(FI), based on the Gini coefficient to measure the fairness of task allocation. The experimental results demonstrate that F3A consistently maintains FI within 0.1 across various cluster sizes and different task quantities, representing an improvement of 76.45% compared to classical fair scheduling algorithms round-robin. F3A exhibits commendable performance in ensuring fairness in task allocation while also demonstrating effectiveness in cost reduction and enhancing user satisfaction.

关键词： Industries Data centers Cloud computing Job shop scheduling Costs Scheduling algorithms Large language models

DCSA: The Deployment Mechanism of Chained Serverless Applications in JointCloud Environment

学校读者我要写书评

暂无评论

DCSA: The Deployment Mechanism of Chained Serverless Applica...

IEEE International Conference on Joint Cloud Computing (JCC)

作者： Yaojie Li Peichang Shi Jianfei Liu Rui Li Fei Gao Penghui Ma Dong Xie Guodong Yi National Key Laboratory of Parallel and Distributed Processing College of Computer Science National University of Defense Technology Changsha China Xiangjiang Lab Changsha China School of Advanced Interdisciplinary Studies Hunan University Of Technology and Business Changsha China

ISBN: (数字)9798350387339

ISBN: (纸本)9798350387346

Serverless computing, comprised of Function as a Service (FaaS) and Backend as a Service (BaaS), has garnered widespread attention owning to its features such as maintenance-free operations, pay-per-use pricing, and automatic scalability. However, practical usage encounters several challenges: 1) The diversity of user applications makes comprehensive performance evaluation difficult, as benchmark and application tests only reflect performance under specific conditions and cannot fully capture users’ actual experiences across different serverless platforms. 2) Disparities in performance and costs across different serverless platforms make it challenging to achieve optimal performance and cost efficiency through single-cloud deployment, thereby underutilizing the advantages of each platform. 3) Vendor lock-in issues restrict the migration of user applications and exacerbate dependence on a single cloud *** address these challenges, this paper proposes a collaborative mechanism, referred to as DCSA, which integrates FaaS and storage services to achieve automatic cross-cloud deployment of user applications while considering both performance and cost comprehensively. Firstly, we adapt the interfaces of different serverless platforms, effectively reducing the complexity of cross-cloud deployment. Secondly, we develop cost and latency models for the cross-cloud deployment of chained serverless applications and propose a deployment scheduling algorithm that simultaneously considers both latency and cost. Finally, we conduct experiments to evaluate the performance of the proposed algorithm. Results demonstrate that our method can effectively reduce latency (up to 2.3%) and lower costs (up to 9.9%).

关键词： Performance evaluation Adaptation models Costs Scheduling algorithms Scalability Serverless computing Collaboration

A distributed Computing Framework Based on Lightweight Variance Reduction Method to Accelerate Machine Learning Training on Blockchain

学校读者我要写书评

暂无评论

China Communications 2020年第9期17卷 77-89页

作者： Zhen Huang Feng Liu Mingxing Tang Jinyan Qiu Yuxing Peng Science and Technology on Parallel and Distributed Laboratory National University of Defense TechnologyChangsha 410000China H.R.Support Center PLABeijing 100000China

To security support large-scale intelligent applications,distributed machine learning based on blockchain is an intuitive solution ***,the distributed machine learning is difficult to train due to that the corresponding optimization solver algorithms converge slowly,which highly demand on computing and memory *** overcome the challenges,we propose a distributed computing framework for L-BFGS optimization algorithm based on variance reduction method,which is a lightweight,few additional cost and parallelized scheme for the model training *** validate the claims,we have conducted several experiments on multiple classical *** show that our proposed computing framework can steadily accelerate the training process of solver in either local mode or distributed mode.

关键词： machine learning optimization algorithm blockchain distributed computing variance reduction