Create, Train, and Deploy a Model

Description

The models.get() and models.create() methods enable you to get an existing model or create and deploy a new model.

Syntax

Use the models.get() method to get an existing model:

my_model = project.models.get('my_model')

Or, the create() method to create and train a new model:

my_model = project.models.create (
    name = 'my_model',
    predict = 'target',
    query = my_table
)

Please note that in the case of LLM models, the parameters can be stored in options. Here is the syntax to create an OpenAI model:

sentiment_classifier = project.models.create (
      name='sentiment_classifier',
      engine='openai', # alternatively: engine=server.ml_engines.openai
      predict='sentiment',
      options={
          'prompt_template':'answer this question: {{questions}}',
          'model_name':'gpt4'
      }
)

Alternatively, you can skip options and define parameters as arguments.

sentiment_classifier = project.models.create (
      name='sentiment_classifier',
      engine='openai', # alternatively: engine=server.ml_engines.openai
      predict='sentiment',
      prompt_template='answer this question: {{questions}}',
      model_name='gpt4'
)

And in the case of time-series model, the additional options are stored in timeseries_options. Here is the syntax to create a time-series model:

ts_model = project.models.create (
     name='time_series_model',
     predict='target',
     query=my_table,
     timeseries_options={
            'order': 'order_date',
            'group': 'category',
            'window': 30,
            'horizon': 4
          }
)