Implementation:SeldonIO Seldon core Seldon Model CRD

Property	Value
Implementation Name	Seldon_Model_CRD
Type	Pattern Doc
Overview	Concrete pattern for declaring Seldon Core 2 Model resources via Kubernetes YAML manifests.
Implements Principle	SeldonIO_Seldon_core_Model_Resource_Definition
Workflow	Model_Deployment
Domains	MLOps, Kubernetes
Source	samples/models/sklearn1.yaml:L1-8
External Dependencies	Kubernetes API (mlops.seldon.io/v1alpha1), rclone
Last Updated	2026-02-13 00:00 GMT

Description

This implementation provides the concrete YAML pattern for declaring Seldon Core 2 Model resources as Kubernetes Custom Resources. The Model CRD manifest is the primary interface between ML practitioners and the Seldon Core 2 platform. It specifies the model artifact location, runtime requirements, and optional resource constraints in a declarative format that the Seldon scheduler consumes.

Code Reference

Source: samples/models/sklearn1.yaml:L1-8

apiVersion: mlops.seldon.io/v1alpha1
kind: Model
metadata:
  name: iris
spec:
  storageUri: "gs://seldon-models/mlserver/iris"
  requirements:
  - sklearn

Key Parameters

Parameter	Type	Description
apiVersion	string	API group and version: mlops.seldon.io/v1alpha1
kind	string	Resource kind: Model
metadata.name	string	Unique model name within the namespace (used for inference routing)
spec.storageUri	string	URI pointing to the model artifact location (supports gs://, s3://, minio://, local paths)
spec.requirements	list[string]	Runtime capability tags that must match a Server (e.g., ["sklearn"], ["tensorflow"], ["huggingface"])
spec.memory	string (optional)	Memory allocation hint for the scheduler (e.g., "100Ki", "1Gi")
spec.server	string (optional)	Pin the model to a specific Server by name

I/O Contract

Inputs

Input	Type	Description
Model artifact	Remote storage object	Serialized model file(s) stored at the URI specified by spec.storageUri (GCS, S3, MinIO, or local path)

Outputs

Output	Type	Description
Kubernetes Model resource	CRD manifest	A declarative resource manifest ready for submission via kubectl apply or seldon model load

Usage Examples

Minimal sklearn Model

apiVersion: mlops.seldon.io/v1alpha1
kind: Model
metadata:
  name: iris
spec:
  storageUri: "gs://seldon-models/mlserver/iris"
  requirements:
  - sklearn

Model with Memory Allocation

apiVersion: mlops.seldon.io/v1alpha1
kind: Model
metadata:
  name: iris-large
spec:
  storageUri: "gs://seldon-models/mlserver/iris"
  requirements:
  - sklearn
  memory: 500Ki

TensorFlow Model

apiVersion: mlops.seldon.io/v1alpha1
kind: Model
metadata:
  name: cifar10
spec:
  storageUri: "gs://seldon-models/triton/cifar10"
  requirements:
  - tensorflow

Applying the Manifest

# Apply using kubectl
kubectl apply -f samples/models/sklearn1.yaml

# Or load using the Seldon CLI
seldon model load -f samples/models/sklearn1.yaml

Knowledge Sources

• Repository: https://github.com/SeldonIO/seldon-core
• Documentation: https://docs.seldon.io/projects/seldon-core/en/v2/

Related Pages

• SeldonIO_Seldon_core_Seldon_Model_CRD implements SeldonIO_Seldon_core_Model_Resource_Definition
• SeldonIO_Seldon_core_Sklearn_Pipeline_Train_And_Serialize produces artifacts for SeldonIO_Seldon_core_Seldon_Model_CRD
• SeldonIO_Seldon_core_Seldon_Model_Load submits SeldonIO_Seldon_core_Seldon_Model_CRD
• Environment:SeldonIO_Seldon_core_Kubernetes_Cluster_Environment
• Environment:SeldonIO_Seldon_core_Go_Build_Toolchain_Environment
• Heuristic:SeldonIO_Seldon_core_Model_Scheduling_Preference_Tip
• Heuristic:SeldonIO_Seldon_core_Autoscaling_Dual_Config_Tip