AN 998: Intel® FPGA AI Suite Inference on Quantized Graphs
ID
786296
Date
12/01/2023
Public
1. Intel® FPGA AI Suite Inference on Quantized Graphs
The Intel® FPGA AI Suite supports running inference on INT8 symmetrically quantized CNNs that are represented in the OpenVINO™ intermediate representation (IR) format. In this example, you will use OpenVINO™ notebooks to quantize a floating point 32-bit ResNet-18 model through the Neural Network Compression Framework (NNCF).
The Intel® FPGA AI Suite provides scripts to set up the environment for OpenVINO™ notebooks and NNCF. It also provides a script to download and organize the data set on which you run inference.