XGBoost Server

If you have a trained XGBoost model saved you can deploy it simply using Seldon’s prepackaged XGBoost server.


  • The model must be named model.bst

  • You must save your model using bst.save_model(file_path)

  • The model is loaded with xgb.Booster(model_file=model_file)

  • Dependencies (otherwise it may not work):

    • numpy == 1.15.4

    • xgboost == 1.2.0

An example for a saved Iris prediction model:

apiVersion: machinelearning.seldon.io/v1alpha2
kind: SeldonDeployment
  name: xgboost
  name: iris
  - graph:
      children: []
      implementation: XGBOOST_SERVER
      modelUri: gs://seldon-models/xgboost/iris
      name: classifier
    name: default
    replicas: 1

Try out a worked notebook