Overview

AMD Ryzen AI Software is a comprehensive suite of tools and runtime libraries designed to facilitate the optimization and deployment of AI inference on AMD Ryzen AI-powered PCs. It enables developers to efficiently port pre-trained models from frameworks like PyTorch or TensorFlow to run on the integrated Neural Processing Unit (NPU) or integrated Graphics Processing Unit (iGPU) found in select Ryzen AI-powered laptops. Applications are deployed using ONNX Runtime, allowing for seamless inference across different hardware configurations.

Features and Capabilities

• Model Deployment:
  • Supports deployment of pre-trained models from PyTorch and TensorFlow.
  • Utilizes ONNX Runtime for cross-platform compatibility.
• Hardware Acceleration:
  • Leverages the integrated NPU for AI-specific tasks, enhancing performance and efficiency.
  • Employs the iGPU for general-purpose computation and machine learning tasks.
• Model Optimization:
  • Provides tools for model quantization and optimization to improve inference speed and reduce resource consumption.
  • Supports INT8 and BF16 configurations for model deployment.
• Developer Tools:
  • Includes a growing pre-trained model zoo for quick experimentation.
  • Offers optimized Large Language Models (LLMs) using ONNX Runtime.
• AI Workload Support:
  • Addresses various AI workloads, including Large Language Models, image and video generation, recommendation systems, and computer vision tasks.
• Execution Modes:
  • Supports hybrid execution modes, utilizing both NPU and iGPU for optimal performance.
  • Offers NPU-only execution modes for compute-intensive operations.
• Open-Source Projects:
  • Encourages exploration of open-source tools to empower developers in analyzing, optimizing, and deploying AI models efficiently across diverse hardware.