The application of Transformer-based large models has achieved numerous success in recent years. However, the exponential growth in the parameters of large models introduces formidable memory challenge for edge deploy...
详细信息
ISBN:
(纸本)9798350380415;9798350380408
The application of Transformer-based large models has achieved numerous success in recent years. However, the exponential growth in the parameters of large models introduces formidable memory challenge for edge deployment. Prior works to address this challenge mainly focus on optimizing the model structure and adopting memory swapping methods. However, the former reduces the inference accuracy, and the latter raises the inference latency. This paper introduces PIPELOAD, a novel memory-efficient pipeline execution mechanism. It reduces memory usage by incorporating dynamic memory management and minimizes inference latency by employing parallel model loading. Based on PIPELOAD mechanism, we present Hermes, a framework optimized for large model inference on edge devices. We evaluate Hermes on Transformer-based models of different sizes. Our experiments illustrate that Hermes achieves up to 4.24x increase in inference speed and 86.7% lower memory consumption than the state-of-the-art pipeline mechanism for BERT and ViT models, 2.58x increase in inference speed and 90.3% lower memory consumption for GPT-style models.
With the surge of end devices and intelligent services, computing resource has begun to migrate from the cloud to end devices to meet the growing demand of users, forming a hierarchical resource distribution pattern i...
详细信息
Clock synchronization is a critical aspect in the operation of Wireless sensor Networks (WSNs), ensuring that sensor nodes maintain accurate time for coordinated communication and data processing. Various clock synchr...
详细信息
ISBN:
(纸本)9798331505264;9798331505271
Clock synchronization is a critical aspect in the operation of Wireless sensor Networks (WSNs), ensuring that sensor nodes maintain accurate time for coordinated communication and data processing. Various clock synchronization algorithms have been developed to address the synchronization challenges in WSNs, aiming to align the clocks of distributed nodes efficiently. These algorithms play a crucial role in achieving synchronization by compensating for clock drift, which can significantly impact the accuracy of timekeeping among sensor nodes. Timing constraints are fundamental in WSNs, as they influence network performance and data transmission reliability. Addressing timing constraints is essential for optimizing network operations and ensuring timely data delivery. In this paper we analysed various synchronization protocols, Algorithm and techniques. Finally, we compared features of TPSN and RBS protocols, based on accuracy, robustness TPSN is best one.
This paper presents a decentralised signal classification approach for data acquired using Internet of Things (IoT) wearable sensors. Traditionally, data from IoT sensors are processed in a centralised fashion, and in...
详细信息
ISBN:
(纸本)9798350300246
This paper presents a decentralised signal classification approach for data acquired using Internet of Things (IoT) wearable sensors. Traditionally, data from IoT sensors are processed in a centralised fashion, and in a single node. This approach has several limitations, such as high energy consumption on the edge sensor, longer response times, etc. We present a distributed processing approach for convolutional neural network (CNN) based classifiers where a single CNN model can be split into multiple sub-networks using early exits. To reduce the transfer of large feature maps between sub-networks, we introduced an encoder-decoder pair at the exit points. Processing of inputs that can be classified with high confidence at an exit point will be terminated early, without needing to traverse the entire network. The initial sub-networks can be deployed on the edge to reduce sensor energy consumption and overall complexity. We also experimented with multiple exit point locations and show that the point of exit can be adjusted for trade-offs between complexity and performance. The proposed system can achieve a sensitivity of 98.45% and an accuracy of 97.55% for electrocardiogram (ECG) classification and save 60% of the data transmitted wirelessly while reducing 38.45% of the complexity.
External and internal disruptions into power system stability motivates to research on a qualified grid resilience system. With this, indispensability of electricity infrastructure emphasizes the critical need for pow...
详细信息
This paper explores an intelligent heavy truck solution with integrated sensing, communication, computation, and control (ISCC), capabilities based on the "vehicle-energy-road-cloud" framework for enclosed i...
详细信息
ISBN:
(纸本)9798350374520;9798350374513
This paper explores an intelligent heavy truck solution with integrated sensing, communication, computation, and control (ISCC), capabilities based on the "vehicle-energy-road-cloud" framework for enclosed intermodal railwayroad transport parks. Enabled by communication-sensing convergence networks, the system achieves ubiquitous connectivity across Multi-Agent individual truck intelligence, vehicle coordination intelligence, edge computing, and cloud platforms. The multi-modal sensor fusion of vehicle-energy-road elements ensures comprehensive environmental cognition by combining truck-mounted devices, energy facilities, and roadside sensors. This allows collaborative decisionmaking through distributed in-vehicle and cloud-based analytics. Through intelligent dispatching, precise control commands guide trucks to safely and efficiently complete loading, weighing, vehicle interworking, and other intermodal transport tasks. This human-in-the-loop framework synergizes sensing, communication, computation, and control to fully unlock the potential of new energy heavy trucks, enhancing the safety, accuracy, and efficiency of freight haulage operations in complex enclosed parks.
The grid integrated with distributed energy resources or generators (DERs) called microgrids are being used extensively because of power hunger in the world. Since the integration of more distributed generations, powe...
详细信息
Collaborative learning is an emerging field of machine learning. In this framework, multiple learning algorithms try to learn from a distributed database. The main idea is to improve the performance of each algorithm ...
详细信息
Development of environments for distributedsystems is a tedious and time-consuming iterative process. The reproducibility of such environments is a crucial factor for rigorous scientific contributions. We think that ...
详细信息
ISBN:
(数字)9781665498562
ISBN:
(纸本)9781665498562
Development of environments for distributedsystems is a tedious and time-consuming iterative process. The reproducibility of such environments is a crucial factor for rigorous scientific contributions. We think that being able to smoothly test environments both locally and on a target distributed platform makes development cycles faster and reduces the friction to adopt better experimental practices. To address this issue, this paper introduces the notion of environment transposition and implements it in NixOS Compose, a tool that generates reproducible distributed environments. It enables users to deploy their environments on virtualized (docker, QEMU) or physical (Grid'5000) platforms with the same unique description of the environment. We show that NixOS Compose enables to build reproducible environments without overhead by comparing it to state-of-the-art solutions for the generation of distributed environments (EnOSlib and Kameleon). NixOS Compose actually enables substantial performance improvements on image building time over Kameleon (up to 11x faster for initial builds and up to 19x faster when building a variation of an existing environment).
There has been a significant change in the management and operation of buildings as a result of the incorporation of distributed Intelligence into Building Management systems. This paper analyzes the revolutionary pot...
详细信息
暂无评论