This section describes how to deploy the quantized ONNX model on the Edge board.
In Vitis AI 3.5, the ONNX Runtime Vitis AI Execution Provider (Vitis AI EP) is provided to hardware-accelerated AI inference with the DPU . It allows you to directly run the quantized ONNX model on the target board. The current Vitis AI EP inside the ONNX Runtime enables acceleration of the neural network model inference using embedded devices such as Zynq UltraScale+ MPSoCs, Versal devices, Versal AI Edge devices, and Kria cards.
The Vitis AI ONNX Runtime Engine (VOE) is the implementation library of Vitis AI EP.
Figure 1. ONNX Runtime Overview