Implementation:SeldonIO Seldon core Seldon Model CRD Explainer Deploy

Field	Value
Type	Pattern Doc
Overview	Concrete pattern for declaring and deploying explainer models in Seldon Core 2.
Domains	MLOps, Explainability, Kubernetes
Workflow	Model_Explainability
Related Principle	SeldonIO_Seldon_core_Explainer_Model_Deployment
Source	samples/models/income-explainer.yaml:L1-9, samples/models/moviesentiment-explainer.yaml:L1-9, samples/explainers/explainers.yaml:L1-26
Last Updated	2026-02-13 00:00 GMT

Code Reference

Income Explainer (AnchorTabular with modelRef)

Source: samples/models/income-explainer.yaml:L1-9

apiVersion: mlops.seldon.io/v1alpha1
kind: Model
metadata:
  name: income-explainer
spec:
  storageUri: "gs://seldon-models/scv2/samples/mlserver_1.6.0/income-sklearn/anchor-explainer"
  explainer:
    type: anchor_tabular
    modelRef: income

Movie Sentiment Explainer (AnchorText with modelRef)

Source: samples/models/moviesentiment-explainer.yaml:L1-9

apiVersion: mlops.seldon.io/v1alpha1
kind: Model
metadata:
  name: sentiment-explainer
spec:
  storageUri: "gs://seldon-models/scv2/samples/mlserver_1.6.0/moviesentiment/anchor-explainer"
  explainer:
    type: anchor_text
    modelRef: sentiment

Key Parameters

Parameter	Description	Values
spec.explainer.type	Alibi explanation algorithm	anchor_tabular, anchor_text, kernel_shap
spec.explainer.modelRef	Name of the base model to explain	e.g., income, sentiment
spec.explainer.pipelineRef	Name of the pipeline to explain (alternative to modelRef)	e.g., sentiment-explain
spec.storageUri	URI to the serialized explainer artifact	e.g., gs://seldon-models/scv2/samples/mlserver_1.6.0/income-sklearn/anchor-explainer

I/O Contract

Inputs

• Explainer CRD YAML: A Model CRD manifest with the spec.explainer section specifying the algorithm type and base model reference
• Base model deployed and ready: The model referenced by modelRef must be in ModelAvailable state

Outputs

• Explainer model loaded on MLServer: The explainer is served by the Alibi-Explain runtime, linked to the base model via modelRef for black-box inference during explanation generation

External Dependencies

• Kubernetes API
• MLServer Alibi-Explain runtime
• Seldon scheduler
• seldon CLI

Usage Examples

Deploying the Income Explainer

# Ensure the base model is loaded first
seldon model load -f samples/models/income.yaml
seldon model status income -w ModelAvailable

# Deploy the explainer
seldon model load -f samples/models/income-explainer.yaml
seldon model status income-explainer -w ModelAvailable

Deploying the Sentiment Explainer

# Ensure the base model is loaded first
seldon model load -f samples/models/moviesentiment.yaml
seldon model status sentiment -w ModelAvailable

# Deploy the explainer
seldon model load -f samples/models/moviesentiment-explainer.yaml
seldon model status sentiment-explainer -w ModelAvailable

Explainer CRD Structure

The explainer CRD extends the standard Model CRD with the spec.explainer section:

apiVersion: mlops.seldon.io/v1alpha1
kind: Model
metadata:
  name: <explainer-name>
spec:
  storageUri: "<uri-to-explainer-artifact>"
  explainer:
    type: <anchor_tabular|anchor_text|kernel_shap>
    modelRef: <base-model-name>       # OR
    pipelineRef: <pipeline-name>      # for pipeline-level explanation

Knowledge Sources

• Repo: SeldonIO/seldon-core
• Doc: Seldon Core 2 Documentation

Related Pages

• SeldonIO_Seldon_core_Explainer_Model_Deployment - parent principle - Declaring and deploying explainer models.
• SeldonIO_Seldon_core_Seldon_Model_Load_For_Explainer - prerequisite - Loading the base model before deploying the explainer.
• SeldonIO_Seldon_core_Seldon_Model_Status_Explainer - next step - Verifying the explainer is ready.
• SeldonIO_Seldon_core_Seldon_Pipeline_CRD_Explainer - alternative - Using pipelineRef for pipeline-level explanation.