Fitted predictor

Looking for usage documentation? Check out Relational predictions.

`FittedPredictor`

A fitted model and its tuning results for one training context.

FittedPredictor(
    context: PredictiveContext,
    _model: RelArenaModel,
    _cache: CacheConfig,
    trials: list[TrialResult] | None,
    config: dict[str, Any],
)

Fields

Name	Type	Default	Description
`context`	`PredictiveContext`	Required	—
`trials`	`list[TrialResult] \| None`	Required	—
`config`	`dict[str, Any]`	Required	—

`FittedPredictor.predict`

View source

Predict original entity IDs with features frozen at the context cutoff.

FittedPredictor.predict(
    query: PredictiveQuery,
    *,
    cache_dir: str | Path | None = None,
) -> pd.DataFrame

Parameters

Parameter	Type	Default	Description
`query`	`PredictiveQuery`	Required	The entities and anchor timestamp to predict.
`cache_dir`	`str \| Path \| None`	`None`	Feature cache; `None` reuses the one passed to `fit`.

Returns

Type	Description
`pd.DataFrame`	One row per entity: the time and entity columns and `<target_col>_pred`.

Reference

tabpfn

tabpfn-client

tabpfn-extensions

tabpfn-rel

REST API

`FittedPredictor`

`FittedPredictor.predict`

​FittedPredictor

​FittedPredictor.predict

`FittedPredictor`

`FittedPredictor.predict`