Qualcomm AI software "family bucket" is here, developed for mobile phones, computers, XR cars and other platforms, netizens: Amazing

Xiaoxiao is from Aofei Temple
Quantum Bit | Public Account QbitAI

Qualcomm’s AI software “family bucket” is finally here.

This "family bucket" includes various toolkits developed by Qualcomm AI, supporting AI frameworks such as TensorFlow and PyTorch , and supporting mainstream operating systems such as Android, Windows, and Linux .

The most important thing is that it is universal for all Snapdragon platforms and is not specifically targeted at mobile phones, computers or a certain type of device.

You know, although Qualcomm has developed a lot of AI toolboxes (SDKs) before , including those for developing automotive, Internet of Things, XR and other platforms, these toolboxes are still independent of each other.

If developers want to create an AI model based on the Qualcomm platform, after writing one for a mobile phone, they have to write another one for a computer or car, or make modifications based on the platform.

It can be said that this latest product directly reduces the difficulty of porting AI models on various Qualcomm platforms.

Integrates Qualcomm AI capabilities

The full name of this product is Qualcomm AI Stack , which is mainly composed of four modules: AI framework, developer library and services, system software and operating system.

Specifically, it supports developers to directly use the AI ​​software stack to develop algorithms based on three AI frameworks: PyTorch, TensorFlow, and ONNX, and these algorithms can be accelerated by the Qualcomm AI engine.

There are no restrictions on the operating system. In addition to the three most mainstream ones, it also supports operating systems such as Ubuntu and CentOS.

Once the AI ​​model is developed, it can be deployed on the Internet of Things, robots and even the cloud in addition to smartphones, computers, cars and XR.

It can be seen that Qualcomm's AI software stack basically integrates the results of Qualcomm's efforts in AI research and development in recent years, including open source tools such as quantization (AIMET) , neural network architecture search (NAS) , and AI development graphical user interface (GUI) .

Qualcomm also specifically mentioned the Google Cloud Vertex AI NAS announced with Google Cloud. This tool has now been integrated into the software stack and can be called at any time.

So, where exactly can this AI software stack be used?

For example, a developer has developed a facial recognition feature that runs on a Qualcomm chip phone, but now he wants to migrate it to other devices such as cars or security cameras.

With the AI ​​software stack, developers can directly deploy this facial recognition function without having to make modifications based on platform differences, and avoid repeated development.

Have you tried this AI software stack?

Reference link:
https://www.qualcomm.com/products/technology/artificial-intelligence/ai-stack