Compile a Model

The Model Compiler is the SiMa.ai toolchain that turns a trained model into an artifact the Neat runtime can execute on the MLSoC. The vision/ONNX path uses the afe Python package to check compatibility, prepare the graph, quantize, compile, and validate accuracy and performance.

Choose your starting point

• Use a precompiled modelStart with a Model Zoo artifact when a suitable model already exists.
• Compile a vision / ONNX modelPrepare, quantize, compile, and validate with the Model Compiler. Start with the first-model walkthrough.
• Compile a GenAI / LLiMa modelUse the LLiMa toolchain for LLMs and other generative models.