Creating custom models

Developers frequently need to create a model that is not using a single framework. This mix and match makes model development easier, but it complicates depending on the model.

Verta supports custom models defined as a class that can perform arbitrary computations and this tutorial shows you how to create one.

Defining a class model

A Class Model must implement this interface:

class MyModel(object):
    def __init__(self, artifacts):
        pass

    def predict(self, data):
        pass

__init__() takes an argument called artifacts, which are covered in this tutorial.

predict() takes one argument, which is the model input data and returns the model output. The model input can be of arbitrary type as defined in DeployedModel.predict(), but it must be serializable to json.

The Verta infrastructure automatically converts known types (arrays, dataframes, and many others) to their json-compatible format automatically.

In order to use this model, you must also register this model with Verta.

Saving a class model

Any class within the Verta platform that has a log_model() method, such as an ExperimentRun, automatically supports logging custom models. They will be serialized in the right way for consumption downstream.

For example, when using an experiment run, you can just do

run.log_model(MyModel)  # class, not instance

and that class will be logged into the system.

The same can be done using a RegisteredModelVersion:

model_version.log_model(MyModel)  # class, not instance

When an object of that class is requested, our tools will automatically build one with the right configuration for you!

More complex models

Not every model is self-contained in its definition. We also support adding runtime artifacts to models.