Ryzen AI-Specific Tutorials#
AMD Ryzen AI is AMD’s NPU-enabled platform for on-device AI inference. Unlike general-purpose quantization, Ryzen AI-specific quantization requires that the model and its quantization configuration satisfy the hardware constraints of the Ryzen AI NPU — such as supported operators, data types, and graph topology. Each tutorial below covers how to meet those constraints for a specific model or use case.
Quantize and deploy ResNet50 on AMD Ryzen AI hardware using Quark for ONNX.
Quantize YOLOv8 object detection model and run it on AMD Ryzen AI hardware.
Use Auto Search with a custom evaluator to find optimal quantization config for MobileNetv2-50 on Ryzen AI.
Automatically search for the best quantization strategy for ResNet50 targeting Ryzen AI.
Apply Auto Search to find the optimal mixed-precision config for YOLOv8 on Ryzen AI.