This project demonstrates using ONNX-MLIR to compile the resnet50 model from the ONNX model zoo, and then leverages BentoML to deploy for inference.
Follow the install instructions. This will include installing prerequisites, including LLVM. Since ONNX-MLIR is a recent project as of this writing, we suggest pulling in the most recent stable version.
Alternatively, docker images are available. See the links on the ONNX-MLIR project page.
See resnet notebook