Qualcomm Brings Its Best-in-Class AI Software Offerings into a Single Package Called The Qualcomm AI Stack That Allows Developers To Take AI models From One Device Type to Another

To enable businesses to create consistent artificial intelligence (AI)-based apps across diverse product form factors, Qualcomm has announced the availability of its AI Stack portfolio for developers. The launch essentially broadens the company’s developer offerings in AI that it already had. Still, Qualcomm claimed that with the introduction of the AI stack, the software’s scope would effectively include all product categories for which it offers hardware, including phones, cars, the internet of things (IoT), and mixed reality (XR), and cloud services.

Utilizing current developer tools and bringing them to additional form factors would be a crucial component of the company’s AI stack. The AI Stack, for example, would now let developers creating an AI application for a Qualcomm-powered smartphone trickle over the same functionality to a car’s infotainment interface or linked home devices.

The company has also made it clear that the AI Stack will support a variety of AI runtime environments, including Meta Platforms’ PyTorch and Google’s open-source TensorFlow for artificial intelligence (AI) and machine learning (ML). To aid developers building on Qualcomm’s AI Stack in integrating the broadest possible variety of capabilities into their AI offering, it will also enable third-party developer libraries, open-source system software, tools, and services.

What is the Qualcomm AI Stack?

To enable OEMs and Developers to build, improve, and deploy AI applications on Qualcomm Technologies’ products while taking full advantage of the speed and effectiveness of the Qualcomm® AI Engine, QTI created this stack.

To make it simple to deploy any AI feature created for one device on another, the AI Stack comprises underpinning developer libraries and services, system software, tools, and compilers. TensorFlow, PyTorch, and runtimes like TFLite and ONNXRT are just a few of the well-known AI frameworks supported by Qualcomm AI Stack. Several tools have been included, including the current Qualcomm® Neural Processing SDK, the well-liked Qualcomm® AI Model Efficiency Toolkit (AIMET) Pro, the AIMET Model Zoo, model analyzers, and Neural Architecture Search (NAS). Additionally, Qualcomm has expanded the Qualcomm AI Engine direct, a powerful AI runtime, across all Qualcomm Technologies’ products, including the Qualcomm® Cloud AI 100 inference processor. Qualcomm also recently transferred the Qualcomm Neural Processing SDK to Microsoft Windows. By allowing developers to deploy current models directly to the AI accelerators on Qualcomm Technologies’ platforms, this latter improvement assists developers in achieving the write-once-run-anywhere aim QTI has established.

Three domain-specific SDKs for autonomous cars, intelligent multimedia SDK for robots and IoT, and virtual reality (Snapdragon SpacesTM SDK) are built on the Qualcomm AI Software Stack. Once more, the fact that these SDKs are based on a single foundation enables developers to support the Qualcomm Cloud AI 100 and the whole array of QTI hardware implementations.

Developers may extend their AI applications to a greater variety of product form factors and types with a product stack like this. The change may make it possible for AI apps built on Qualcomm hardware processors to operate more fluidly on Windows and sync with other devices running various software systems. Microsoft also stated at Qualcomm’s launch that it will support Qualcomm’s AI Stack for Windows developers, including Qualcomm’s own Neural Processing SDK for Windows.

Conclusions

The need for a unified software stack has grown as QTI extends its reach into the connected, intelligent edge. When Qualcomm Technologies unifies its wide variety, it becomes easier for developers to create apps using the same tools regardless of the intended deployment platform, including mobile, automotive, XR, compute, IoT, and cloud platforms of AI software. The business has now taken on this issue head-on with Qualcomm AI Stack and will be able to provide additional AI development tools and other domain-specific SDKs in the future.

References:

• https://www.qualcomm.com/news/onq/2022/06/Qualcomm_unveils_Qualcomm_AI_Stack
• https://www.theregister.com/2022/06/23/qualcomm_ai_devs_software/

Please Don't Forget To Join Our ML Subreddit