Intel AI Platforms Accelerate Microsoft Phi-3 GenAI Models

[btarunr]

Intel has validated and optimized its AI product portfolio across client, edge and data center for several of Microsoft's Phi-3 family of open models. The Phi-3 family of small, open models can run on lower-compute hardware, be more easily fine-tuned to meet specific requirements and enable developers to build applications that run locally. Intel's supported products include Intel Gaudi AI accelerators and Intel Xeon processors for data center applications and Intel Core Ultra processors and Intel Arc graphics for client.

"We provide customers and developers with powerful AI solutions that utilize the industry's latest AI models and software. Our active collaboration with fellow leaders in the AI software ecosystem, like Microsoft, is key to bringing AI everywhere. We're proud to work closely with Microsoft to ensure Intel hardware - spanning data center, edge and client - actively supports several new Phi-3 models," said Pallavi Mahajan, Intel corporate vice president and general manager, Data Center and AI Software.



As part of its mission to bring AI everywhere, Intel continuously invests in the AI software ecosystem by collaborating with AI leaders and innovators.

Intel worked with Microsoft to enable Phi-3 model support for its central processing units (CPUs), graphics processing units (GPUs) and Intel Gaudi accelerators on launch day. Intel also co-designed the accelerator abstraction in DeepSpeed, which is an easy-to-use deep learning optimization software suite, and extended the automatic tensor parallelism support for Phi-3 and other models on Hugging Face.

The size of Phi-3 models is well-suited to be used for on-device inference and makes lightweight model development like fine-tuning or customization on AI PCs and edge devices possible. Intel client hardware is accelerated through comprehensive software frameworks and tools, including PyTorch and Intel Extension for PyTorch used for local research and development and OpenVINO Toolkit for model deployment and inference.

What's Next: Intel is committed to meet the generative AI needs of its enterprise customers and will continue to support and optimize software for Phi-3 and other leading state-of-the-art language models.

For performance and technical details, visit the Intel Developer Blog.

View at TechPowerUp Main Site

 

[JWNoctis]

FWIW the newly-released 4k-context medium version model is actually pretty nice*; Probably one of the best open-weight LLM you can run right now with 16GB of RAM or VRAM, quantized.

You don't need anything as specific as the "Intel AI platform" for it to run well, though. A GPU with 16GB of VRAM would do just fine, and do it much faster - It should still be acceptably fast on mostly any modern CPU though, should anyone wish to try.

*Reasonable-sounding as long as you don't ask it for anything specific, specialized, or niche, as always for these sort of thing, but especially for smaller ones runnable on a consumer video card, or at usable speed in RAM.