Xilinx is releasing its Version 1.4 of its Vitis AI inference acceleration software platform, which includes new solution stack support for the company’s 7nm Versal adaptive computing acceleration platforms and its 16nm-based Kria adaptive system-on-modules portfolio.
The Vitis platform was first announced almost two years ago to help product designers bring AI to their products, but in recent months more semiconductor companies have been addressing the growing need for more AI inference support and capabilities as AI technology has spread to even more products. Earlier this week, Nvidia made its own software announcement on the AI inference front, announcing Version 8 of its Tensor RT platform, with improved performance and accuracy for AI inference.
For Xilinx’s part, a company blog post today stated that Vitis 1.4 will now support the Versal AI Core Series VCK190 Evaluation Kit and the VCK5000 Versal Development Card for AI Inference. The VCK190 kit, which allows designers to develop solutions using AI and DSP engines capable of delivering over 100x greater compute performance than today's server-class CPUs, is ideal for “supporting high throughput AI inference and signal processing applications from cloud to edge,” the blog post stated.
The VCK5000, an out-of-the-box solution, also targets designs requiring high throughput AI inference and signal processing compute performance.
“Together with the Kria KV260 Vision AI Starter Kit, these new AI development platforms provide more possibilities for users to achieve superior AI inference performance, scalability and cloud-to-edge deployment options in AI productization,” the blog post stated.
For customers that don’t want to customize their own neural network models, Xilinx can help them realize the benefits of highly-efficient AI inference acceleration with the company’s own set of free AI models, which are optimized, retrainable and deployable by anyone. “In Vitis AI 1.4, the diversity of this AI Model Zoo has been increased to include state-of-the-art models for 4D radar detection, image-lidar sensor fusion, Surround view 3D detection, depth estimation, super resolution and many more, totaling hundreds of models from different ML frameworks,” the blog post stated.
Xilinx said it is aiming to help improve AI productization efforts and enable the creation of domain-specific architectures (DSA). To demonstrate the high-efficiency of DSA in AI inference acceleration, Xilinx submitted a ResNet50 closed-division benchmark in the MLPerf Inference v1.0 results earlier this year.
“The results measured how quickly a trained neural network can process new data for a wide range of applications on a variety of form factors. We achieved a result of 5,921 images per second (img/s) using the Versal VCK5000 PCI-E card. It outperformed the acceleration performance result achieved by a Nvidia T4 card under the same mode,” the blog post stated.
Regarding further enhancements, Xilinx said that in Vitis AI 1.4, the quantizer, optimizer and compiler tools now support popular machine learning frameworks Pytorch, Tensorflow 1.x, Tensorflow 2.x and Caffe. Also, new APIs and operator features in the platform allow more AI models deployment across multiple devices.
RELATED: Xilinx unveils Vitis software for AI designers