lenskit.pipeline.RecPipelineBuilder#

class lenskit.pipeline.RecPipelineBuilder#

Bases: object

Builder to help assemble common pipeline patterns.

This is a convenience class; you can always directly assemble a Pipeline if this class’s behavior is inadquate.

Methods

`__init__`()
`build`([name])	Build the specified pipeline.
`candidate_selector`(sel[, config])	Specify the candidate selector component.
`predicts_ratings`(*[, transform, fallback])	Specify that this pipeline will predict ratings, optionally providing a rating transformer and fallback scorer for the rating predictions.
`ranker`([rank, config, n])	Specify the ranker to use.
`reranker`(reranker[, config])	Specify a reranker to add to the pipeline.
`scorer`(score[, config])	Specify the scoring model.

Attributes

is_predictor

scorer(score, config=None)#

Specify the scoring model.

Parameters:

ranker(rank=None, config=None, *, n=None)#

Specify the ranker to use. If None, sets up a TopNRanker with n=n.

Parameters:

candidate_selector(sel, config=None)#

Specify the candidate selector component. The component should accept a query as its input and return an item list.

Parameters:

predicts_ratings(*, transform=None, fallback=None)#

Specify that this pipeline will predict ratings, optionally providing a rating transformer and fallback scorer for the rating predictions.

Parameters:

transform (Component | None) – A component to transform scores prior to returning them. If supplied, it will be applied to both primary scores and fallback scores.
fallback (Component | None) – A fallback scorer to use when the primary scorer cannot score an item. The fallback should accept query and items inputs, and return an item list.

build(name=None)#

Build the specified pipeline.

reranker(reranker, config=None)#

Specify a reranker to add to the pipeline.

Parameters:

reranker (Component | ComponentConstructor) – The reranker to use in the pipeline.
config (object | None) – Configuration parameters to initialize the reranker if a reranker is specified.