Jun Yang

Join us on Thursday, Oct 17, for our webinar on "Supercharging AI Platforms with Anyscale" 🚀 ⚡ Discover how Anyscale can optimize AI platforms with key capabilities including: 🔵Advanced Instance Management: Launch clusters in <60s, Scale to 1000s of nodes, and rely on Anyscale's intelligent instance selection that searches across on-demand and spot instances to reduce costs. 🔵Global Compute Management: Efficiently manage resources across cloud, on-prem, and Kubernetes. 🔵Preventing AI Sprawl: Anyscale's Governance Suite provides fine grained control to set quotas, monitor usage, and the observability teams need to optimize AI workloads. Don’t miss it—sign up today: https://lnkd.in/gFpWMB3C

13

We encourage you to read the July TCAS-II paper highlight, "A 28nm 8Kb Reconfigurable SRAM Computing-In-Memory Macro With Input-Sparsity Optimized DTC for Multi-Mode MAC Operations". 📝 Authored by: Kanglin Xiao; Xin Qiao; Xiaoxin Cui; Jiahao Song; Haoyang Luo; Xin’an Wang; Yuan Wang Computing-in-memory (CIM) is an emerging approach for alleviating von Neumann bottleneck and improving energy efficiency and throughput. This brief presents a reconfigurable SRAM CIM macro supporting multi-mode multiply-and-accumulate (MAC) operations, including binary weight network (BWN) MAC, ternary weight network (TWN) MAC, and multi-bit MAC operations. 🔗 Read more on IEEE Xplore here: https://loom.ly/7CbDOMw #TCASII #PaperHighlight #CircuitsandSystems #ResearchBriefs

4

We encourage you to delve into the recently published paper "StrideHD: A Binary Hyperdimensional Computing System Utilizing Window Striding for Image Classification". 📝 Authored by: Dehua Liang; Jun Shiomi; Noriyuki Miura; Hiromitsu Awano IEEE Open Journal of Circuits and Systems (Volume: 5) In this work, the authors propose a novel HDC system: StrideHD. The uniqueness of StrideHD is to capture the critical features utilizing window striding method and organize the HDC architecture in a distributed way. 🔗 Read more on IEEE Xplore here: https://loom.ly/oTorzVM #OJCAS #OA #OpenAccess #PaperHighlight

1

Meet AGUVIS: A pure vision-based framework for autonomous Computer/Mobile-use GUI agents, operating seamlessly across web, desktop, and mobile platforms without UI code. Key Features & Contributions - Pure Vision Framework: First fully autonomous pure vision GUI agent capable of performing tasks independently without relying on closed-source models - Cross-Platform Unification: Unified action space and plugin system that works consistently across different GUI environments - Comprehensive Dataset: Large-scale dataset of GUI agent trajectories with multimodal grounding and reasoning - Two-Stage Training: Novel training pipeline focusing on GUI grounding followed by planning and reasoning - Inner Monologue: Explicit planning and reasoning capabilities integrated into the model training - Generalization to OOD (Out-of-Distribution): AGUVIS demonstrates impressive generalization on OSWorld benchmark even without desktop trajectory data during training. Project Page: https://lnkd.in/guzkwpbB Paper: https://lnkd.in/gs7wXyJS GitHub: https://lnkd.in/gWeSkFmK Collaboratiion among Yiheng Xu Zekun Wang Junli Wang Dunjie Lu Tianbao Xie Amrita Saha Doyen Sahoo Tao Yu Caiming Xiong

155

6 Comments

📢 New User Experiment Blog: July 2024 Out Now! 🖥️🔬 Exciting news! Our latest User Experiment blog is now live, featuring groundbreaking research on Multi-Tiered Memory Management by Prof. Dong Li and his team at UC Merced. Highlights: • Innovative approach to managing large, multi-tiered memory systems • Real-world experiment using Chameleon's four-tier memory nodes • Insights on overcoming challenges in HPC memory profiling and migration Dive into the world of high-performance computing and discover how Chameleon is powering cutting-edge research. Read the full blog post here: https://lnkd.in/gzU4EQh5 #ChameleonCloud #UserExperiment #HPC #ResearchComputing

6

Most major tech companies are at Computex in Taiwan this week, and we're seeing a flood of announcements. The industry is buzzing following Jensen Huang's announcement of the next-generation Rubin GPU platform and associated Vera Arm CPUs. These will replace Grace, Hopper, and Grace Hopper as NVIDIA's scalable AI platform in 2026 and beyond. It's excellent to see NVIDIA honor another woman in science with their platform names, in this case American astronomer Vera Rubin who passed away in 2016. #Rundown #Computex #AI

1

🤔 Curious about how SynapseAI's Graph Compiler works with #Intel #Gaudi? In our latest blog, we explore the impact of Intel Gaudi’s Graph Compiler on large language model (LLM) inference, discussing both its strengths and weaknesses. 💡 Discover how #vLLM performs on #Gaudi_2 compared to #NVIDIA_A100 in our analysis! 🔎 Kudos to Taesu Kim for all the effort! 👏

13

Huida's next-generation AI chip was revealed to be delayed by three months before mass production. Nvidia's next-generation Blackwell architecture AI flagship chip, the GB200, will face a delay of at least three months before mass production due to design defects. An undisclosed Microsoft employee indicated that Nvidia informed Microsoft of the shipment delays last week. This delay affects major tech companies, including Google, which ordered over 400,000 chips with a value exceeding $10 billion, and Meta, which also placed a $10 billion order. The GB200 chip includes two interconnected Blackwell GPUs and a Grace CPU. The design flaw was identified by TSMC engineers during preparations for mass production, resulting in the need for design adjustments and new trial production. Initially expected to begin mass production in the third quarter and shipments in the fourth quarter, the timeline has now shifted to mass production in the fourth quarter and shipments postponed to the first quarter of the following year. #AI #Taiwan https://lnkd.in/gYXSTeqg

1

Check out "Car-Following-Response-Based Vehicle Classification via Deep Learning," by Tianyi Li and Raphael Stern in the inaugural issue of ACM Journal on Autonomous Transportation Systems (JATS): https://bit.ly/3V4H0U0 #autonomousvehicle #deeplearning #appliedcomputing #engineering

1

Yesterday’s biweekly teatime seminar presented by Postdoctoral Researcher, Dr. Shiqing Li, was a stimulating discussion on how to accelerate sparse matrix operations using embedded FPGAs,   Emerging applications in Internet of Things (IoT) and embedded systems such as autonomous driving require efficient and low-power hardware accelerators to make real-time decisions and responses. Coupled with the exponential growth of data used in these applications, high-performance computing platforms are crucial to support these applications. FPGAs are used to fill this gap because they provide flexibility to design customized hardware accelerators, which can improve performance compared to traditional CPUs. These are exciting times ahead as we harness this potential to drive innovation!

2

🌟 Exploring the Intersection of Business and Technology re AI! 🌟 Before my decade-long focus on AI, connecting technology and business in various domains has always been my passion; This is why I enjoyed this conversation with David Luan - No BS, No Twitter exaggerations, No clickbate - Just a super well informed, super smart David sharing his insights from years of hands-on experience: #Startegy, #Software, #Hardware, #Platforms, #ModelSpace and everything in between - LMK what you think of his projections. In This Episode: 1. The Biggest Lessons from OpenAI and Google Brain: What did OpenAI realize that no one else did that allowed them to steal the show with ChatGPT? Why did it take 6 years after the introduction of transformers for ChatGPT to be released? What are 1-2 of David’s biggest lessons from his time leading teams at OpenAI and Google Brain? 2. Foundation Models: The Hard Truths: Why does David strongly disagree that the performance of foundation models is at a stage of diminishing returns? Why does David believe there will only be 5-7 foundation model providers? What will separate those who win vs those who do not? Does David believe we are seeing the commoditization of foundation models? How and when will we solve core problems of both reasoning and memory for foundation models? 3. Bunding vs Unbundling: Why Chips Are Coming for Models: Why does David believe that Jensen and Nvidia have to move into the model layer to sustain their competitive advantage? Why does David believe that the largest model providers have to make their own chips to make their business model sustainable? What does David believe is the future of the chip and infrastructure layer? 4. The Application Layer: Why Everyone Will Have an Agent: What is the difference between traditional RPA vs agents? Why are agents a 1,000x larger business than RPA? In a world where everyone has an agent, what does the future of work look like? Why does David disagree with the notion of “selling the work” and not the tool? What is the business model for the next generation of application-layer AI companies? https://lnkd.in/dUpWWzvN

4

Smaller #genAI models are changing the game! Introducing #mistral-NeMo by #nvidia and #mistralAI with 128k token context window - a tool for businesses that can run on RTX GPUs which you may have at your home. Super inexpensive with no high cloud costs, keeps data private on your private computer! #nvidia #mistralai #genaiforenterprise #responsibleai

10

It's finally here 🎉 We’re excited to kick off our series of technical analyses, in which we're exploring the performance implications of NVIDIA's H200 GPU for the use case of LLM, VLM, and DiT inference and training. In the first installment, we measure the multi-node inference overhead and explore the benefits of FP8 quantization. 🛠️ Hardware: NVIDIA H200 The H200 shows great promise, particularly for extremely large Transformers models, as it allows to handle BFLOAT16 without multi-node inference, removing any inter-node communication overhead. 🐋 Model: DeepSeek V3 We chose DeepSeek V3 as our case study model due to its SOTA model capabilities and high demanding GPU resources for model serving in production and research environments. 🚀 Serving framework: SGLang Although we plan to benchmark other inference systems such as vLLM and TensorRT-LLM, we landed on SGLang due to its performance-oriented design and easy-to-modify Python code base. This first blog in the series is a part of our ongoing co-research with SGLang, for which we provided GPU infrastructure and engineering hours. Read it in full and see the results: https://lnkd.in/ePt_b3gm

32

1 Comment

We're pumped to welcome Chip Huyen of Voltron Data to the PyTorch Conference 2024 keynote stage! Explore the list of keynote speakers: https://hubs.la/Q02JLPJb0 Register & join us September 18-19 in San Francisco! https://hubs.la/Q02JLF6T0 #PyTorchConf #PyTorch #PyTorchFoundation #AI #MachineLearning #ML #DeepLearning #OpenSource #OpenSourceSoftware #OpenSourceDevelopment #OpenSourceCommunity #OSS #LinuxFoundation #events #linux #SoftwareSupplyChainNews

53

We encourage you to read the recently published article "TM-GAN: A Transformer-Based Multi-Modal Generative Adversarial Network for Guided Depth Image Super-Resolution". Authored by: Jiang Zhu; Van Kwan Zhi Koh; Zhiping Lin; Bihan Wen 📝 Volume: 14, Issue: 2, June 2024 Motivation: Convolutional neural network (CNN) has limitations in capturing global information effectively, and their traditional regression training techniques can sometimes lead to challenges in the precise generating of high-frequency details, unlike transformers that have shown remarkable success in deep learning through the self-attention mechanism. Contribution: We propose a novel transformer-based multimodal generative adversarial network (TM-GAN) for GDSR applications and the proposed TM-GAN learns non-local correlation matrices of the target and guidance images that are used to guide the multi-modality information fusion in the corresponding feature space, thereby enhancing the quality of target image reconstruction. Results: Utilizing TM-GAN for guided depth super-resolution, our approach consistently outperforms recent and widely recognized methods, achieving state-of-the-art results across various benchmark datasets. 🔗 Continue reading on IEEE Xplore here: https://loom.ly/Emvcovs #JETCAS #CircuitsandSystems #graphicalabstract #ReadMore

7