Skip to content

Latest commit

 

History

History
53 lines (34 loc) · 2.02 KB

walkthrough.md

File metadata and controls

53 lines (34 loc) · 2.02 KB

Walkthrough

Workbench creation

Go to RHODS dashboard. Login with Openshift. Possible highlights:

  • How to configure Jupyter Notebooks
  • How to create and link persistent and bucket storage

Jupyter notebooks

From the RHODS dashboard, open the jupyter notebooks. Login with Openshift. Possible highlights:

  • Create and run a pipeline with Elyra
  • Create and run a pipeline using KFP tekton sdk
  • Commit your code to the local gitea instance

MLOps automation

Find the gitea route on the gitea namespace. Log in with data-scientist-1 / rhods.

  • Create a new pull request and merge to the prod branch This will trigger the Openshift pipeline and the data science pipeline
  • See the pipeline running on RHODS dashboard
  • Wait for competion
  • Show the tagged model in the object storage (stored under rhods/onnx/model-TAG.onnx)
  • Show the metrics directly on the pull request comment or on the mlflow dashboard (mlflow instance is deployed on mlflow namespace)

Model serving

Go back to rhods dashboard

  • Configuring a model server with route and or authentication
  • Model deployment. It is pointing to the versionned model tagged prod

Live inference

On mnist namespace, find the webapp route and open the frontend. Draw a digit to make a prediction. Optionnaly explain the goal of the serverless function as a pre processor for the HTTP requests made by the frontend.

animation

Camel integration

You can explain the 2 camel integrations. You can highlight:

  • The minio to kafka inteagration (using KamelBinding)
  • The KafkaSource that triggers the serverless function on a new event
  • The kafka to SQL inteagration (using Integration)

Batch inference

Open the Grafana dashboard. The route is on the mnist namespace. Open the Inference dashboard. Log in with admin / rhods

  • Add images from the ./test/batch/images to the images bucket
  • You can also use the ./test/batch/push-minio.sh script but it requires the mb cli from minio
  • On grafana see the images processed in real time