H2OPredict | Supported External Model Types | Teradata Package for Python - H2OPredict - Teradata Package for Python

H2OPredict performs a prediction on each row of the input table using a model previously trained in H2O and then loaded into the database. The model uses an interchange format called MOJO and it is loaded as a blob to a table in Teradata database by the user.

The following are examples of H2OPredict() function call.

Example Setup

• Import necessary modules.

  >>> import os, teradataml

  >>> from teradataml.options.configure import configure

  >>> from teradataml import H2Oredict, DataFrame, load_example_data, save_byom, retrieve_byom

• Load example data.

  >>> load_example_data("byom", "iris_input")

• Create teradataml DataFrame object.

  >>> iris_test = DataFrame("iris_input")

• Set install location of the BYOM functions.

  >>> # Set install location of BYOM functions.
  >>> configure.byom_install_location = "mldb"

Example 1: Run a query with GLM model and overwrite cached models

The query also includes arguments model_type, enable_options and model_output_fields.

• Load model file into Vantage.

  >>> model_file = os.path.join(os.path.dirname(teradataml.__file__), "data", "models", "iris_mojo_glm_h2o_model")

• Save the model.

  >>> save_byom("iris_mojo_glm_h2o_model", model_file, "byom_models")

• Retrieve the model.

  >>> modeldata = retrieve_byom("iris_mojo_glm_h2o_model", table_name="byom_models")

• Pass the output of the retrieve_model API as an input to the PMMLPredict function to score data.

  >>> result = H2OPredict(newdata=iris_test,
                          newdata_partition_column='id',
                          newdata_order_column='id',
                          modeldata=modeldata,
                          modeldata_order_column='model_id',
                          model_output_fields=['label', 'classProbabilities'],
                          accumulate=['id', 'sepal_length', 'petal_length'],
                          overwrite_cached_models='*',
                          enable_options='stageProbabilities',
                          model_type='OpenSource')

• Print the results.

  print(result.result)

Example 2: Run a query with XGBoost model and overwrite cached models

The query also includes arguments model_type, enable_options and model_output_fields.

• Load model file into Vantage.

  >>> model_file = os.path.join(os.path.dirname(teradataml.__file__), "data", "models", "iris_mojo_xgb_h2o_model")

• Save the model.

  >>> save_byom("iris_mojo_xgb_h2o_model", model_file, "byom_models")

• Retrieve the model.

  >>> modeldata = retrieve_byom("iris_mojo_xgb_h2o_model", table_name="byom_models")

• Pass the output of the retrieve_model API as an input to the PMMLPredict function to score data.

  >>> result = H2OPredict(newdata=iris_test,
                          newdata_partition_column='id',
                          newdata_order_column='id',
                          modeldata=modeldata,
                          modeldata_order_column='model_id',
                          model_output_fields=['label', 'classProbabilities'],
                          accumulate=['id', 'sepal_length', 'petal_length'],
                          overwrite_cached_models='*',
                          enable_options='stageProbabilities',
                          model_type='OpenSource')

• Print the results.

  print(result.result)

Example 3: Run a query with a licensed model

This example runs a query with a licensed model with id 'licensed_model1' from the table 'byom_licensed_models' and associated license key stored in column 'license_key' of the table 'license' present in the schema 'mldb'.

• Retrieve model.

  modeldata = retrieve_byom('licensed_model1',
                             table_name='byom_licensed_models',
                             license='license_key',
                             is_license_column=True,
                             license_table_name='license',
                             license_schema_name='mldb')

• Pass the output of the retrieve_model API as an input to the H2OPredict function to score data.

  result = H2OPredict(newdata=iris_test,
                      newdata_partition_column='id',
                      newdata_order_column='id',
                      modeldata=modeldata,
                      modeldata_order_column='model_id',
                      model_output_fields=['label', 'classProbabilities'],
                      accumulate=['id', 'sepal_length', 'petal_length'],
                      overwrite_cached_models='*',
                      enable_options='stageProbabilities',
                      model_type='OpenSource')

• Print the results.

  print(result.result)