python3_sklearn_runtime_params

Python Sklearn Inference Model with Runtime Parameters (example)

This model is intended to work with the Python 3 Scikit-Learn Drop-In Environment. The supplied pkl file is a scikit-learn model trained on juniors_3_year_stats_regression.csv with a Grade 2014 as the target (regression), though any binary or regression model trained using the libraries outlined in Python 3 Scikit-Learn Drop-In Environment will work.

In this example, the custom.py file contains a dummy transform() function that demonstrates how to access runtime parameter values set in a Custom Inference Model in the DataRobot platform.

⚠️ Runtime Parameters is currently a Beta Feature that will need to be enabled by your administrator.

Instructions

This example contains a model-metadata.yaml file with a runtimeParameterDefinitions section where you must declare the runtime parameters. The custom.py script uses the helper functions from the datarobot_drum.RuntimeParameters class, available in datarobot-drum package version 1.10 and newer (included in the latest pre-built drop-in environments).

Create a new custom model with these files and the Python Drop-In Environment. Ensure that you upload the model-metadata.yaml file to the top-level of the model's file structure. After you add the metadata file, DataRobot reads the parameter definitions and updates the Assembly tab UI to display the default values, allowing you to override them and provide any required credentials.

Run locally with drum

Create a values file to set overrides for the parameters defined in this example. An example could look as follows:

# option1 will use its default value (ABCD 123) if you don't override it below
#option1: Hello World

# set a value for option2
option2: goodbye!

# option3 has no default and if we don't override it below, it will simply have an implicit
# default value of `None`.
#option3: NULL

# Credential type params are stored as a mapping. The `credentialType` key will always be present
# and the value will inform what specific credential type is being injected. The other key/values
# vary with credential type:
#   https://docs.datarobot.com/en/docs/api/reference/public-api/credentials.html#properties_3
encryption_key:
  credentialType: rsa
  rsaPrivateKey: xfzRmbBoIT/89/QdACH0f8PI6Idq55JOnkzy9nkMXu2uyt3VHZWm6krbLtInBvc+

Paths are relative to ./datarobot-user-models:

drum score --code-dir model_templates/python3_sklearn_runtime_params --target-type regression --input tests/testdata/juniors_3_year_stats_regression.csv --runtime-params-file <path_to_values_file>