wrap_mlmodel (Python agent API)

Syntax

newrelic.agent.wrap_mlmodel(model, name=None, version=None, feature_names=None, label_names=None, metadata=None)

Enables manual instrumentation of machine learning models.

Requirements

Python agent version 9.1.0 or higher.

Description

This allows for manual instrumentation of machine learning models.

Parameters

Parameter	Description
`model` object	Required. `model` is the fitted machine learning model.
`name` string	Optional. The name of the custom model.
`version` string	Optional. The release version of the custom model.
`feature_names` list of string	Optional. A list of strings denoting the feature name(s).
`label_names` list of string	Optional. A list of strings denoting the label name(s).
`metadata` dict	Optional. Metadata to attach to the model.

Return values

None.

Examples

Wrap machine learning model

An example of instrumenting a custom machine learning model:

def wrap_ml_example():
    x_train = [[0, 0], [1, 1]]
    y_train = [0, 1]
    x_test = [[1.0, 2.0]]

    model = CustomTestModel().fit(x_train, y_train)
    wrap_mlmodel(
        model,
        name="MyCustomModel",
        version="1.2.3",
        feature=["feature0", "feature1"],
        label=["label0"],
        metadata={"metadata1": "value1", "metadata2": "value2"},
    )

    labels = model.predict(x_test)

    return model