Serving systems for Large Language Models (LLMs) are often optimized to improve quality of service (QoS) and throughput. However, due to the lack of open-sourced LLM serving workloads, these systems are frequently evaluated under unrealistic workload assumptions. Consequently, performance may degrade when these systems are deployed in real-world scenarios. This work presents BurstGPT, an LLM serving workload with 5.29 million traces from regional Azure OpenAI GPT services over 121 days. BurstGPT captures realistic LLM serving characteristics through detailed tracing of: (1) Concurrency of requests: It traces burstiness variations of requests in Azure OpenAI GPT services, revealing diversified concurrency patterns in different services and model types. (2) Response Lengths of requests: It traces the auto-regressive serving processes of GPT models, showing statistical relations between requests and their responses. (3) Failures of requests: It traces failures of conversation and API services, showing intensive resource needs and limited resource availability of such services in Azure. Details of the characteristics can serve multiple purposes in LLM serving optimizations, such as system evaluation and trace provisioning. In our demo evaluation with BurstGPT, we observe that frequent variations in BurstGPT reveal declines in efficiency, stability, or reliability in realistic LLM serving. We identify that the generalization of KV cache management and request scheduling optimization is not guaranteed for different workloads, especially when systems are poorly optimized for unrealistic workloads. We have made the dataset publicly available to encourage further research at //github.com/HPMLL/BurstGPT.
相關內容
The volume of unlabelled Earth observation (EO) data is huge, but many important applications lack labelled training data. However, EO data offers the unique opportunity to pair data from different modalities and sensors automatically based on geographic location and time, at virtually no human labor cost. We seize this opportunity to create MMEarth, a diverse multi-modal pretraining dataset at global scale. Using this new corpus of 1.2 million locations, we propose a Multi-Pretext Masked Autoencoder (MP-MAE) approach to learn general-purpose representations for optical satellite images. Our approach builds on the ConvNeXt V2 architecture, a fully convolutional masked autoencoder (MAE). Drawing upon a suite of multi-modal pretext tasks, we demonstrate that our MP-MAE approach outperforms both MAEs pretrained on ImageNet and MAEs pretrained on domain-specific satellite images. This is shown on several downstream tasks including image classification and semantic segmentation. We find that pretraining with multi-modal pretext tasks notably improves the linear probing performance compared to pretraining on optical satellite images only. This also leads to better label efficiency and parameter efficiency which are crucial aspects in global scale applications.
General purpose AI, such as ChatGPT, seems to have lowered the barriers for the public to use AI and harness its power. However, the governance and development of AI still remain in the hands of a few, and the pace of development is accelerating without a comprehensive assessment of risks. As a first step towards democratic risk assessment and design of general purpose AI, we introduce PARTICIP-AI, a carefully designed framework for laypeople to speculate and assess AI use cases and their impacts. Our framework allows us to study more nuanced and detailed public opinions on AI through collecting use cases, surfacing diverse harms through risk assessment under alternate scenarios (i.e., developing and not developing a use case), and illuminating tensions over AI development through making a concluding choice on its development. To showcase the promise of our framework towards informing democratic AI development, we run a medium-scale study with inputs from 295 demographically diverse participants. Our analyses show that participants' responses emphasize applications for personal life and society, contrasting with most current AI development's business focus. We also surface diverse set of envisioned harms such as distrust in AI and institutions, complementary to those defined by experts. Furthermore, we found that perceived impact of not developing use cases significantly predicted participants' judgements of whether AI use cases should be developed, and highlighted lay users' concerns of techno-solutionism. We conclude with a discussion on how frameworks like PARTICIP-AI can further guide democratic AI development and governance.
AI-based systems leverage recent advances in the field of AI/ML by combining traditional software systems with AI components. Applications are increasingly being developed in this way. Software engineers can usually rely on a plethora of supporting information on how to use and implement any given technology. For AI-based systems, however, such information is scarce. Specifically, guidance on how to securely design the architecture is not available to the extent as for other systems. We present 16 architectural security guidelines for the design of AI-based systems that were curated via a multi-vocal literature review. The guidelines could support practitioners with actionable advice on the secure development of AI-based systems. Further, we mapped the guidelines to typical components of AI-based systems and observed a high coverage where 6 out of 8 generic components have at least one guideline associated to them.
Vision Transformers (ViTs) have shown impressive performance and have become a unified backbone for multiple vision tasks. However, both the attention mechanism and multi-layer perceptrons (MLPs) in ViTs are not sufficiently efficient due to dense multiplications, leading to costly training and inference. To this end, we propose to reparameterize pre-trained ViTs with a mixture of multiplication primitives, e.g., bitwise shifts and additions, towards a new type of multiplication-reduced model, dubbed $\textbf{ShiftAddViT}$, which aims to achieve end-to-end inference speedups on GPUs without requiring training from scratch. Specifically, all $\texttt{MatMuls}$ among queries, keys, and values are reparameterized using additive kernels, after mapping queries and keys to binary codes in Hamming space. The remaining MLPs or linear layers are then reparameterized with shift kernels. We utilize TVM to implement and optimize those customized kernels for practical hardware deployment on GPUs. We find that such a reparameterization on attention maintains model accuracy, while inevitably leading to accuracy drops when being applied to MLPs. To marry the best of both worlds, we further propose a new mixture of experts (MoE) framework to reparameterize MLPs by taking multiplication or its primitives as experts, e.g., multiplication and shift, and designing a new latency-aware load-balancing loss. Such a loss helps to train a generic router for assigning a dynamic amount of input tokens to different experts according to their latency. Extensive experiments on various 2D/3D Transformer-based vision tasks consistently validate the effectiveness of our proposed ShiftAddViT, achieving up to $\textbf{5.18$\times$}$ latency reductions on GPUs and $\textbf{42.9}$% energy savings, while maintaining a comparable accuracy as original or efficient ViTs.
Vision Transformers (ViTs) represent a groundbreaking shift in machine learning approaches to computer vision. Unlike traditional approaches, ViTs employ the self-attention mechanism, which has been widely used in natural language processing, to analyze image patches. Despite their advantages in modeling visual tasks, deploying ViTs on hardware platforms, notably Field-Programmable Gate Arrays (FPGAs), introduces considerable challenges. These challenges stem primarily from the non-linear calculations and high computational and memory demands of ViTs. This paper introduces CHOSEN, a software-hardware co-design framework to address these challenges and offer an automated framework for ViT deployment on the FPGAs in order to maximize performance. Our framework is built upon three fundamental contributions: multi-kernel design to maximize the bandwidth, mainly targeting benefits of multi DDR memory banks, approximate non-linear functions that exhibit minimal accuracy degradation, and efficient use of available logic blocks on the FPGA, and efficient compiler to maximize the performance and memory-efficiency of the computing kernels by presenting a novel algorithm for design space exploration to find optimal hardware configuration that achieves optimal throughput and latency. Compared to the state-of-the-art ViT accelerators, CHOSEN achieves a 1.5x and 1.42x improvement in the throughput on the DeiT-S and DeiT-B models.
The design of energy-efficient, high-performance, and reliable Convolutional Neural Network (CNN) accelerators involves significant challenges due to complex power and thermal management issues. This paper introduces SAfEPaTh, a novel system-level approach for accurately estimating power and temperature in tile-based CNN accelerators. By addressing both steady-state and transient-state scenarios, SAfEPaTh effectively captures the dynamic effects of pipeline bubbles in interlayer pipelines, utilizing real CNN workloads for comprehensive evaluation. Unlike traditional methods, it eliminates the need for circuit-level simulations or on-chip measurements. Our methodology leverages TANIA, a cutting-edge hybrid digital-analog tile-based accelerator featuring analog-in-memory computing cores alongside digital cores. Through rigorous simulation results using the ResNet18 model, we demonstrate SAfEPaTh's capability to accurately estimate power and temperature within 500 seconds, encompassing CNN model accelerator mapping exploration and detailed power and thermal estimations. This efficiency and accuracy make SAfEPaTh an invaluable tool for designers, enabling them to optimize performance while adhering to stringent power and thermal constraints. Furthermore, SAfEPaTh's adaptability extends its utility across various CNN models and accelerator architectures, underscoring its broad applicability in the field. This study contributes significantly to the advancement of energy-efficient and reliable CNN accelerator designs, addressing critical challenges in dynamic power and thermal management.
Over time, agriculture is the most consistent activity, and it evolves every day. It contributes to a vast majority of the Gross Domestic Product (GDP) of Nigeria but as ironic as it may be, there is still hunger in significant parts of the country due to low productivity in the agricultural sector and comparison to the geometric population growth. During the first half of 2022, agriculture contributed about 23% of the country's GDP while the industry and services sector had a share of the remaining 77%. This showed that with the high rate of agricultural activities, Nigeria has not achieved food security for the teeming population. and more productivity levels can be attained. Technology can/will assist Nigeria in overcoming global poverty and hunger quicker in both rural and urban areas. Today, there are many types of agricultural technologies available for farmers all over the world to increase productivity. Major technological advancements include indoor vertical farming, automation, robotics, livestock technology, modern greenhouse practices, precision agriculture, artificial intelligence, and blockchain. Mobile phones have one of the highest adoption rates of technologies developed within the last century. Digitalization will bring consumers and farmers closer together to access the shortest supply chain possible and reduce rural poverty and hunger. The paper will review the different agricultural technologies and propose a mobile solution, code Sell Harvest, to make farming more sustainable and secure food. Keywords: Sell Harvest, Agriculture, Technology, Artificial Intelligence, and Digital Farming.
Personalized Federated Learning (pFL) has emerged as a promising solution to tackle data heterogeneity across clients in FL. However, existing pFL methods either (1) introduce high communication and computation costs or (2) overfit to local data, which can be limited in scope, and are vulnerable to evolved test samples with natural shifts. In this paper, we propose PerAda, a parameter-efficient pFL framework that reduces communication and computational costs and exhibits superior generalization performance, especially under test-time distribution shifts. PerAda reduces the costs by leveraging the power of pretrained models and only updates and communicates a small number of additional parameters from adapters. PerAda has good generalization since it regularizes each client's personalized adapter with a global adapter, while the global adapter uses knowledge distillation to aggregate generalized information from all clients. Theoretically, we provide generalization bounds to explain why PerAda improves generalization, and we prove its convergence to stationary points under non-convex settings. Empirically, PerAda demonstrates competitive personalized performance (+4.85% on CheXpert) and enables better out-of-distribution generalization (+5.23% on CIFAR-10-C) on different datasets across natural and medical domains compared with baselines, while only updating 12.6% of parameters per model based on the adapter. Our code is available at //github.com/NVlabs/PerAda.
Sequential recommendation (SR) is to accurately recommend a list of items for a user based on her current accessed ones. While new-coming users continuously arrive in the real world, one crucial task is to have inductive SR that can produce embeddings of users and items without re-training. Given user-item interactions can be extremely sparse, another critical task is to have transferable SR that can transfer the knowledge derived from one domain with rich data to another domain. In this work, we aim to present the holistic SR that simultaneously accommodates conventional, inductive, and transferable settings. We propose a novel deep learning-based model, Relational Temporal Attentive Graph Neural Networks (RetaGNN), for holistic SR. The main idea of RetaGNN is three-fold. First, to have inductive and transferable capabilities, we train a relational attentive GNN on the local subgraph extracted from a user-item pair, in which the learnable weight matrices are on various relations among users, items, and attributes, rather than nodes or edges. Second, long-term and short-term temporal patterns of user preferences are encoded by a proposed sequential self-attention mechanism. Third, a relation-aware regularization term is devised for better training of RetaGNN. Experiments conducted on MovieLens, Instagram, and Book-Crossing datasets exhibit that RetaGNN can outperform state-of-the-art methods under conventional, inductive, and transferable settings. The derived attention weights also bring model explainability.
Conversational recommender systems (CRS) aim to recommend high-quality items to users through interactive conversations. Although several efforts have been made for CRS, two major issues still remain to be solved. First, the conversation data itself lacks of sufficient contextual information for accurately understanding users' preference. Second, there is a semantic gap between natural language expression and item-level user preference. To address these issues, we incorporate both word-oriented and entity-oriented knowledge graphs (KG) to enhance the data representations in CRSs, and adopt Mutual Information Maximization to align the word-level and entity-level semantic spaces. Based on the aligned semantic representations, we further develop a KG-enhanced recommender component for making accurate recommendations, and a KG-enhanced dialog component that can generate informative keywords or entities in the response text. Extensive experiments have demonstrated the effectiveness of our approach in yielding better performance on both recommendation and conversation tasks.
小貼士
登錄享
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191