Developers Building On-Device Models Will Love Qualcomm’s New AI Hub

0
486


image-qualcomm-ai-hub-1

Qualcomm

Last year, we saw a surge of interest and developments in generative AI applications. Now we are in the next era of the technology — with on-device AI becoming more prevalent across our personal devices. At Mobile World Congress, Qualcomm unveiled a tool that simplifies how developers can access and download AI models right to their test devices.

Also: What to expect from MWC 2024: Transparent laptops, AI phones, smart rings, more

The new Qualcomm AI Hub contains a library of over 75 generative AI models that developers can easily download onto Qualcomm-powered devices, with new models continually being added over time.

The library contains some of the most popular models in the industry such as Whisper, OpenAI’s automatic speech recognition (ASR) system, and Stable DIffusion, Stability AI’s text-to-image model. The Qualcomm AI Hub automatically applies hardware-aware optimizations to the models for superior on-device AI performance.

Furthermore, the models are refined to make the best use of all the cores found within the Qualcomm AI Engine, including the NPU, CPU, and GPU, resulting in better power efficiency, lower memory utilization, and up to four times inferencing speeds, according to the company’s Monday release.

The advantage of on-device AI, in addition to overall increased application performance, is that by forgoing any friction with the cloud, there is increased privacy, which is especially important when creating models that use databases with confidential or proprietary data.

Also: Instant evolution: If AI can design a robot in 26 seconds, what else can it do?

“The Qualcomm AI Hub provides developers with a comprehensive AI model library to quickly and easily integrate pre-optimized AI models into their applications, leading to faster, more reliable and private user experiences,” said Durga Malladi, Qualcomm SVP of technology planning and edge solutions.

Qualcomm shares that the optimized models are available today both on the Qualcomm AI Hub, GitHub, and Hugging Face. Developers can sign up for access starting today.