kernels.harness.RetroTask

One unit of work: fit one model to one company x line cohort as of a

Usage

Source

kernels.harness.RetroTask(
    model,
    warehouse,
    line,
    company_code,
    as_of,
    loss_field=None,
    clamp_paid=False,
    seed=None,
    fit_kwargs=dict()
)

training cutoff, score it against the realized outcome.

warehouse must already be concrete (a local path or a pinned github://owner/repo@publish_id with the mart cached) - see data.schedule_p.pinned_source. fit_kwargs carries entry-specific arguments (variant, growth_curve) verbatim: unlike sampler settings they are NOT signature-filtered, so a typo fails loudly as an error row instead of silently fitting the default.

Parameter Attributes

model: str

warehouse: str

line: str

company_code: str

as_of: str

loss_field: str | None = None

clamp_paid: bool = False

seed: int | None = None

fit_kwargs: dict = dict()