grdryn / example-onnx

ODH ai-edge example MLOps project

This project contains some stuff to build the ChRIS spleen-seg model into a runnable image, using the ODH ai-edge MLOps build Pipeline. It is intended to build an image that can be used with veniceofcode/chbox.

How to run this example

• Provision aiedge-e2e Tekton Pipeline. See the aiedge-e2e README.

• Create and populate the test data PVC. See the populate-test-data README for more information.

• Create an image registry secret containing a credential that can be used to push the resulting image to the target image repo, and link it to the pipeline ServiceAccount. See the ODH ai-edge docs on setting up credentials for more information on this.

• [Optional] If your model is stored in S3-compatible object storage, create a secret with a credential that can be used to download that, and link it to the pipeline ServiceAccount. See the ODH ai-edge docs on setting up credentials for more information on this.

• Choose the example PipelineRun that you wish to provision, and use oc create to run that. See the examples README for more information on the examples.

  • Update the target-image-tag-references to point to an image registry repo that you want the pipeline to push to (and which the previously created credential has access to push to).