Edit

Windows ML CLI

Windows ML CLI is a command-line tool for building portable, performant, and high-quality AI models for Windows ML. You can start from a source model on Hugging Face or from your own pipeline, then produce a hardware-optimized artifact in a reproducible workflow.

Windows ML CLI handles conversion, graph optimization, and compilation across AMD, Intel, NVIDIA, and Qualcomm targets. You can also use it in CI/CD pipelines to validate models and prepare them for release.

What you can do

  • Build once, run across hardware - Use primitive commands such as export, analyze, optimize, quantize, and compile, or use winml build to generate a workflow that produces portable models for Windows hardware.
  • Inspect each stage - Review operator compatibility, shape mismatches, graph optimizations, and execution-provider-aware tuning throughout the pipeline.
  • Automate model preparation - Integrate CLI-driven workflows with your existing build and release automation.

What you get out of the box

  • Windows ML execution provider support - Use the same commands across supported execution providers.
  • Curated model catalog - Start from a verified list of supported models that run across Windows ML execution providers.
  • Bring your own ONNX model - Analyze and optimize an existing ONNX model, even if you did not convert it from PyTorch.

Learn more