tabular

agentic_ml.estimators.regression.tabular ¶

Tabular regression: task definition and a convenience researcher.

This is the reference implementation of the :class:~agentic_ml.core.task.Task extension seam. Note how little is task-specific — prompt sections and metrics — while partitioning, evaluation, tools and orchestration are all inherited.

TabularRegressionTask ¶

Bases: Task

A supervised regression problem on tabular data.

TabularRegressionResearcher ¶

TabularRegressionResearcher(model: object | None = None)

Bases: Researcher

A :class:~agentic_ml.research.researcher.Researcher preset for tabular regression.

Pins task to a fresh :class:TabularRegressionTask so callers only need to supply the Strands model. Every attribute otherwise comes straight from the base :class:~agentic_ml.research.researcher.Researcher:

Attributes:

Name	Type	Description
`task`		Always a new :class:`TabularRegressionTask` instance (fixed, not user-provided).
`model`		The Strands model instance driving the agent, or `None` if not yet provided.
`agent`	`object \| None`	The Strands agent built the first time :meth:`~agentic_ml.research.researcher. Researcher.research` runs; `None` before that.

Source code in src/agentic_ml/estimators/regression/tabular.py

def __init__(self, model: object | None = None) -> None:
    super().__init__(TabularRegressionTask(), model=model)