You can store a user-defined Python function for embedded execution.

The following Python code creates a function that receives prediction data. The function loads the trained model from the OML4Py datastore and returns the result by calling the predict method with the prediction data.

You need to register the function for embedded execution with OML by using the oml.script.create method. Note that this function is enclosed within triple quotes.

func = """def error_model_predict_(X):
    import oml
    objs = oml.ds.load('spatial_error_ds', objs=['spatial_error'], to_globals=False)
    error_model = objs['spatial_error']
    pred = error_model.predict(X)
    return pred.tolist()"""
 
oml.script.create("errorModelPredict", func, is_global=True, overwrite=True)

The oml.script.create function adds a user-defined Python function to the OML script repository. The is_global parameter specifies whether to create a global Python function or if it is available only to the current user. The overwrite parameter specifies whether to overwrite the Python function if it already exists.