Seldon Pipeline Canary Promotion

Iris Model

We will:

• Deploy a pretrained sklearn iris model in a Seldon Core V2 Pipeline

• Load test the model

• View request payloads

• Canary a new XGBoost model

• Load test canary model

• Promote the canary model

Deploy Model

Create the model using the wizard with a name of your choice:

URI: gs://seldon-models/scv2/samples/rolling/iris/v1
Type: Seldon ML Pipeline
Runtime: SciKit Learn

All other defaults can be left as provided.

Start Load Test

Once the model is running start a load test with the following request payload:

{
  "inputs": [
    {
      "name": "predict",
      "data": [
        0.38606369295833043,
        0.006894049558299753,
        0.6104082981607108,
        0.3958954239450676
      ],
      "datatype": "FP64",
      "shape": [
        1,
        4
      ]
    }
  ]
}

When running you should see metrics on dashboard. Enter the request logs screen to view requests. If this doesn’t work, consult the metrics or request logging docs section for debugging.

You can also see core metrics from the dashboard.

Create Canary

Create an XGBoost canary model using the saved model at:

URI: gs://seldon-models/xgboost/iris
Type: Seldon ML Pipeline
Runtime: XGBoost

Rerun the load test, and you should see metrics for both default and canary models.

Promote the XGBoost Canary to be the main model using the PROMOTE CANARY button.