Model Optimization

The model optimization flow allows you to perform several key tasks:

• Browse a list of all available AI models and get more information about them.

• Import your AI model.

• Convert models to the Intermediate Representation (IR) format, optimize them, and download them to your Edge Developer Toolbox workspace.

• Perform various types of benchmarking on Intel® platform, including: Comparing the performance of different models when running on the same processor, or the performance of one model when running on different processors.

• Assess a model’s performance when using one of two inference modes - latency optimization which is typically used for lightweight models and real-time services, or throughput optimization which is typically used for models designed for high-performance applications.

You can follow along with this example, where we import two AI models and compare their performance on the Intel® platform.

• Launch Model Optimization and Import an AI Model

• Quantize Model

• Select Target Environment and Benchmarking

• Benchmark Another Model or Target Architecture and Compare Performance

• Use Optimized Model in Edge Developer Toolbox