Vitis AI provided in code form to implement deep learning acceleration
Support for example designs for cloud platforms, endpoints, edge, etc.

Xilinx has launched a free download service for its AI inference development platform, Vitis AI.
Vitis AI is provided in code form to enable software developers to implement deep learning acceleration by combining with the integrated software platform Vitis. It integrates Domain Specific Architecture (DSA) and supports not only optimization for Xilinx hardware but also programming using frameworks such as TensorFlow and Caffe.

It also provides a tool that can optimize, compress, and compile training AI models running on Xilinx devices in as little as one minute. This supports open-source Vitis acceleration libraries and Vitis AI models, as well as related example designs that can be used on cloud platforms and from endpoints to the edge.

Xilinx first unveiled Vitis, an integrated software platform, at the XDF (Xilinx Developer Forum) held in the U.S. last October to enable new developers in various fields, including software engineers and AI scientists, to take advantage of hardware-adaptive solutions.

Vitis is a solution that can automatically customize software or algorithm code to Xilinx hardware architectures without the need for hardware expertise. By connecting with general-purpose software developer tools instead of proprietary development environments and enabling the use of a rich, optimized open-source library, the Vitis platform allows developers to focus solely on developing their own algorithms.

While Xilinx continues to support Vivado Design Suite, a program support tool that utilizes hardware code, it also contributes to improving hardware developer productivity by separately providing the Vitis platform as a hardware module package with software calling capabilities.