Step 5: Deploy a model

Model deployments make trained models available in your environment. Follow these steps to create, activate, and verify model deployments.

Create a deployment

To deploy a model, specify its modelId, deployment name, spaceId, and other configuration details in the request. You also need to set enablePredictions to true to enable predictions for this model in the following API call:

curl -L -X POST "https://<TENANT>/api/v1/ml/deployments" ^
-H "Authorization: Bearer <ACCESS_TOKEN>" ^
-H "Content-Type: application/json" ^
-d "{
    \"data\": {
        \"type\": \"deployment\",
        \"attributes\": {
            \"name\": \"<NAME>\",
            \"spaceId\": \"<SPACE_ID>\",
            \"description\": \"<DESCRIPTION>\",
            \"modelId\": \"<MODEL_ID>\",
            \"enablePredictions\": true,
            \"deprecated\": false
        }
    }
}"

If the deployment is successful, the 201 response includes:

  • the id, which is the deployment ID that you’ll need for activating the model and generating predictions.
  • the modelId, which is the model ID associated with the deployment.

Response example

{
    "data": {
        "type": "deployment",
        "id": "4215f16e-4cc7-432c-bb5d-fc02f16b9c06",
        "attributes": {
            "id": "4215f16e-4cc7-432c-bb5d-fc02f16b9c06",
            "createdAt": "2024-12-01T16:34:31.552381459Z",
            "updatedAt": "2024-12-01T16:34:31.552381459Z",
            "name": "Model deployment for churn predictions",
            "spaceId": "6745f737f536738170dfe82f",
            "description": "Model deployment for churn predictions",
            "modelId": "bed12707-7deb-4a83-99f1-c936aa4a0ebd",
            "enablePredictions": true,
            "deprecated": false,
            "createdBy": "67475097984561d02f0cb3dc",
            "ownerId": "67475097984561d02f0cb3dc",
            "errorMessage": null
        }
    }
}

With your model deployed in your environment, you can now activate it to make it ready to generate predictions.

Activate a model

Once a deployment is created, it needs to be activated to generate predictions. Use the deployment ID from the previous step to activate the model.

You can activate a model with the following API call:

curl -L -X POST "https://<TENANT>/api/v1/ml/deployments/<DEPLOYMENT_ID>/actions/activate-models" ^
-H "Authorization: Bearer <ACCESS_TOKEN>"

When the model is successfully activated, you’ll get a 204 response.

Verify model activation

After activation, verify that the model’s status is updated to enabled. Use the model’s modelId to review its details with the following API call:

curl -L "https://<TENANT>/api/v1/ml/experiments/<EXPERIMENT_ID>/models/<MODEL_ID>" ^
-H "Authorization: Bearer <ACCESS_TOKEN>"

In the response, "modelState": "enabled" means that the model is activated and ready to generate predictions.

Next step

With the model deployment activated, you can now generate predictions in real-time or batch. Predictions allow you to use the deployed model for inference on new data.

Was this page helpful?