Models & Labs
NVIDIA Unveils AI Storage Advancements at FMS
NVIDIA Bloghigh confidence
Why it matters
- →Open sourcing cuFile APIs allows for faster, more secure data access by GPUs.
- →The Storage-Next initiative aims to standardize and optimize AI storage solutions.
- →SCADA framework enhances AI performance by efficiently managing data access.
NVIDIA has announced significant advancements in AI storage at the Future of Memory and Storage conference. The company is open sourcing its cuFile APIs, allowing GPUs to directly access storage, which reduces data access times to microseconds. This initiative is part of a broader effort, including the Storage-Next initiative, to create a secure and efficient storage ecosystem for AI. NVIDIA's SCADA framework further optimizes data access by enabling GPUs to retrieve only the necessary data, enhancing AI performance. These developments aim to improve the productivity and efficiency of AI infrastructure.
Read originalMore from NVIDIA Blog
© NVIDIA BlogMarket & Regulationresearch
NVIDIA Joins NSF AI Hubs Program
NVIDIA's participation in the NSF's State and Regional AI Hubs program represents a pivotal move to enhance AI research and education nationwide. By collaborating with educational institutions and other partners, NVIDIA aims to bolster AI infrastructure and resources, leveraging its successful AI initiative with the University of Florida as a blueprint. This effort will provide shared AI computing resources, which are crucial for accelerating scientific discovery and equipping students with skills for the AI economy. The initiative highlights the need to integrate AI infrastructure with workforce development, aligning with both regional and national priorities.
NVIDIA Blog
© NVIDIA BlogModels & Labsmodels
NVIDIA Alpamayo 2 Super Now Available for AVs
NVIDIA's Alpamayo 2 Super model is now commercially available, marking a significant step forward for autonomous vehicle technology. Built on the Cosmos 3 Super Reasoner, this model excels in handling complex driving scenarios, offering advanced reasoning capabilities crucial for robotaxis and autonomous vehicles. Its open licensing under OpenMDW-1.1 allows developers to adapt and deploy the model without additional permissions, fostering innovation and control over proprietary data. This release not only enhances the AV ecosystem with state-of-the-art reasoning but also democratizes access to cutting-edge AI tools, enabling safer and more efficient autonomous driving solutions.
NVIDIA Blog
© NVIDIA BlogMarket & Regulationresearch
AI Leaders Propose SAFE Guidelines for Cybersecurity
The Open Secure AI Alliance, with over 120 organizations including NVIDIA, Cisco, and Red Hat, is developing the SAFE guidelines to enhance cybersecurity transparency in AI systems. These guidelines aim to transform AI cybersecurity incidents into shared knowledge, promoting collective defense across the ecosystem. By confidentially collecting and analyzing incidents, the initiative seeks to identify recurring failures and provide evidence-based recommendations. This collaborative approach could significantly bolster defenses against emerging threats, fostering a more secure AI landscape.
NVIDIA Blog
More in Models & Labs
Models & Labsmodels
llama.cpp b10278 Release Expands Platform Support
The latest b10278 release of llama.cpp continues its trend of broadening platform compatibility, making it a versatile choice for developers across different systems. Notably, the release includes support for ROCm 7.2 on Ubuntu x64, which is significant for AMD GPU users seeking alternatives to NVIDIA's CUDA. The inclusion of Vulkan support on both Ubuntu and Windows platforms further enhances its appeal for developers working with graphics-intensive applications. While there are no groundbreaking new features, this update solidifies llama.cpp's position as a flexible and inclusive inference runtime for diverse hardware configurations.
llama.cpp Releases
Models & Labsmodels
llama.cpp b10280 Release Expands Platform Support
The b10280 release of llama.cpp marks another step in broadening its reach across various platforms, making it more adaptable for different systems. This update introduces Vulkan support on both Ubuntu and Windows, alongside ROCm 7.2 for Ubuntu, which is a significant boost for AMD GPU users. Windows x64 now benefits from the inclusion of CUDA 12 and 13 DLLs, enhancing its utility for developers. While there are no new models or quantization methods, this release reinforces llama.cpp's role as a flexible and comprehensive solution for AI inference across a wide range of hardware configurations.
llama.cpp Releases
Models & Labsmodels
Llama.cpp b10285 Release Enhances Deepseek-OCR
The latest b10285 release of llama.cpp introduces significant improvements for deepseek-ocr, particularly with multi-row batching support. This update allows for more efficient processing by weaving deepseek-ocr rows in one shot rather than individually, which could enhance performance in OCR tasks. The release also includes a variety of platform-specific builds, such as support for ROCm 7.2 on Ubuntu and CUDA 13 on Windows. While there are no groundbreaking new features, these enhancements make llama.cpp a more versatile tool for developers working with OCR and other AI applications.
llama.cpp Releases