lenskit.pipeline#

LensKit pipeline abstraction.

Submodules#

`components`	Definition of the component interfaces.
`config`	Pydantic models for pipeline configuration and serialization support.
`nodes`	Node objects used to represent pipelines internally. It is very rare

Attributes#

CloneMethod

Classes#

`PipelineBuilder`	Builder for LensKit recommendation pipelines. Pipelines
`PipelineCache`	A cache to share components between pipelines.
`RecPipelineBuilder`	Builder to help assemble common pipeline patterns.
`ComponentInputHook`	Inspect or process data as it passes to a component's input.
`Pipeline`	LensKit recommendation pipeline. This is the core abstraction for using
`PipelineProfiler`	Collect pipeline run statistics for profiling pipeline executions.
`ProfileSink`	Interface for creating recording pipeline run profiling information.
`PipelineState`	Full results of running a pipeline. A pipeline state is a dictionary

Functions#

`predict_pipeline`(scorer, *[, fallback, name])	Create a pipeline that predicts ratings, but does not include any
`topn_pipeline`(scorer[, config, predicts_ratings, n, name])	Create a pipeline that produces top-N recommendations using a scoring model.

Package Contents#

lenskit.pipeline.predict_pipeline(scorer, *, fallback=True, name=None)#

Create a pipeline that predicts ratings, but does not include any ranking capabilities. Mostly userful for testing and historical purposes. The resulting pipeline must be called with an item list.

Stability:

Caller (see Stability Levels).

Parameters:

scorer (lenskit.pipeline.components.Component) – The scorer to use in the pipeline (it will added with the component name scorer, see Component Names).
fallback (bool | lenskit.pipeline.components.Component) – Whether to use a fallback predictor when the scorer cannot score. When configured, the scorer node is the scorer, and the rating-predictor node applies the fallback.
name (str | None) – The pipeline name.

Return type:

lenskit.pipeline._impl.Pipeline

lenskit.pipeline.topn_pipeline(scorer, config=None, *, predicts_ratings=False, n=None, name=None)#

Create a pipeline that produces top-N recommendations using a scoring model.