Come on, Xilinx Vitis AI 1.2 is now available for download!

Latest update time：2020-07-22

Reads：

Click on the blue words to follow us

Vitis AI: Best-in-class AI inference from edge to cloud

Vitis™ AI Development Environment is Xilinx's development platform for AI inference on Xilinx hardware platforms, including edge devices and Alveo cards. It consists of optimized IP, tools, libraries, models, and example designs. Designed with efficiency and ease of use in mind, Vitis AI can fully tap the potential of AI acceleration on Xilinx FPGAs and ACAPs.

The new version 1.2 adds the following features:

Vitis™ AI Quantizer and DNNDK Runtime Open Source
14 new reference models added to AI Model Zoo (Pytorch, Caffe, Tensorflow)
Vitis AI quantizer supports optimized models (pruning)
Updated DPU naming convention to be consistent across all configurations
Introducing Vitis AI Profiler for Edge and Cloud
ONNXRuntime and TVM support Vitis AI
Added support for Alveo U50/U50LV and U280
Alveo U50/U50LV DPU DPUCAHX8H microarchitecture improvements
Updated DPU TRD to support Vitis 2020.1 and Vivado 2020.1
Vitis AI supports Pytorch CNN universal access (beta)

Introduction to the Vitis AI Platform

◆◆ AI Optimizer ◆◆

With world-leading model compression technology, we can reduce model complexity by 5 to 50 times with minimal impact on accuracy. Deep compression takes your AI inference performance to a new level.

◆◆ AI Quantizer ◆◆

By converting 32-bit floating-point weights and activations to fixed-point such as INT8, AI Quantizer can reduce computational complexity without affecting prediction accuracy. Fixed-point network models require less memory bandwidth and are therefore faster and more power efficient than floating-point network models.

◆◆ AI Compiler ◆◆

Maps AI models to efficient instruction sets and data flows. It also performs advanced optimization tasks such as layer fusion and instruction scheduling, and maximizes the reuse of on-chip memory.

◆◆ AI Configurator ◆◆

The performance profiler helps programmers deeply analyze the efficiency and utilization of AI inference implementations.

◆◆ AI Library ◆◆

The runtime provides a series of lightweight C++ and Python APIs that enable convenient application development. In addition, it also provides efficient task scheduling, memory management, and interrupt handling.

Click Read original article to download more Vitis AI technical documents and design resources.