capabilities.work

Windows ML Load and Evaluate Model

A 3-step recipe with Microsoft Windows 10. We can help you get access.

Recipe · 3 steps

Windows ML Load and Evaluate Model

Load an ONNX model, create a session, and evaluate it against input bindings.

  1. Load the ONNX model from the supplied file path and capture the assigned model id and feature counts.
  2. Create an evaluation session for the loaded model bound to the requested device kind.
  3. Evaluate the model with the supplied input bindings and capture the output tensors and whether the evaluation succeeded.
What you'll need to have ready
  • filePath: Path to the ONNX model file to load.
  • deviceKind: Device to use for evaluation.
  • bindings: Map of input feature names to their input values.
  • correlationId: Correlation id for tracking the evaluation.
For your developer

Arazzo workflow microsoft-windows-10:load-and-evaluate-model, 3 operations: loadMLModel → createMLSession → evaluateMLModel. Specs on APIs.io.