LLMAlternatingRecommender

class baybe.recommenders.meta.llm.LLMAlternatingRecommender

Bases: SequentialMetaRecommender

A recommender alternating between an LLM recommender and another one.

A convenience specialization of SequentialMetaRecommender that cycles between a language-model recommender and another recommender, advancing whenever new measurements become available. The sequence must consist of exactly two recommenders, the first of which must be an LLMRecommender.

Public methods

__init__(recommenders[, mode, ...])	Method generated by attrs for class LLMAlternatingRecommender.
from_dict(dictionary)	Create an object from its dictionary representation.
from_json(source, /)	Create an object from its JSON representation.
from_model(model, experiment_description, *)	Create the recommender directly from a LiteLLM model identifier.
get_current_recommender()	Deprecated! Use select_recommender() or get_non_meta_recommender() instead.
get_next_recommender()	Deprecated! Use select_recommender() or get_non_meta_recommender() instead.
get_non_meta_recommender([batch_size, ...])	Follow the meta recommender chain to the selected non-meta recommender.
recommend(batch_size, searchspace[, ...])	See baybe.recommenders.base.RecommenderProtocol.recommend().
select_recommender([batch_size, ...])	Select a recommender for the given experimentation context.
to_dict()	Create an object's dictionary representation.
to_json([sink, overwrite])	Create an object's JSON representation.

Public attributes and properties

is_stateful	Boolean indicating if the meta recommender is stateful.
mode	Defines what shall happen when the last recommender in the sequence has been consumed but additional recommender changes are triggered:
recommenders	A finite-length sequence of recommenders to be used.

__init__(recommenders: Iterable, mode: Literal['raise', 'reuse_last', 'cyclic'] = 'raise', *, _n_last_measurements: int = 0)

Method generated by attrs for class LLMAlternatingRecommender.

For details on the parameters, see Public attributes and properties.

classmethod from_dict(dictionary: dict)

Create an object from its dictionary representation.

Parameters:

dictionary (dict) – The dictionary representation.

Return type:

TypeVar(_T, bound= SerialMixin)

Returns:

The reconstructed object.

classmethod from_json(source: str | Path | SupportsRead[str], /)

Create an object from its JSON representation.

Parameters:

source (Union[str, Path, SupportsRead[str]]) –

The JSON source. Can be:

• A string containing JSON content.

• A file path or Path object pointing to a JSON file.

• A file-like object with a read() method.

Raises:

ValueError – If source is not one of the allowed types.

Return type:

TypeVar(_T, bound= SerialMixin)

Returns:

The reconstructed object.

classmethod from_model(model: str, experiment_description: str, *, litellm_args: dict[str, Any] | None = None, recommender: RecommenderProtocol | None = None)

Create the recommender directly from a LiteLLM model identifier.

Parameters:

• model (str) – The LiteLLM model identifier for the LLM recommender.

• experiment_description (str) – Textual description of the experiment for the LLM.

• litellm_args (Optional[dict[str, Any]]) – Optional additional arguments passed to LiteLLM.

• recommender (Optional[RecommenderProtocol]) – The recommender alternated with the LLM. Defaults to a BotorchRecommender.

Return type:

LLMAlternatingRecommender

Returns:

The configured recommender, cycling between the LLM and recommender.

get_current_recommender()

Deprecated! Use select_recommender() or get_non_meta_recommender() instead.

Return type:

PureRecommender

get_next_recommender()

Deprecated! Use select_recommender() or get_non_meta_recommender() instead.

Return type:

PureRecommender

get_non_meta_recommender(batch_size: int | None = None, searchspace: SearchSpace | None = None, objective: Objective | None = None, measurements: DataFrame | None = None, pending_experiments: DataFrame | None = None)

Follow the meta recommender chain to the selected non-meta recommender.

Recursively calls select_recommender() until a non-meta recommender is encountered, which is then returned. Effectively, this extracts the recommender responsible for generating the recommendations for the specified context.

See baybe.recommenders.base.RecommenderProtocol.recommend() for details on the method arguments.

Return type:

RecommenderProtocol

recommend(batch_size: int, searchspace: SearchSpace, objective: Objective | None = None, measurements: DataFrame | None = None, pending_experiments: DataFrame | None = None)

See baybe.recommenders.base.RecommenderProtocol.recommend().

Return type:

DataFrame

select_recommender(batch_size: int | None = None, searchspace: SearchSpace | None = None, objective: Objective | None = None, measurements: DataFrame | None = None, pending_experiments: DataFrame | None = None)

Select a recommender for the given experimentation context.

See baybe.recommenders.base.RecommenderProtocol.recommend() for details on the method arguments.

Return type:

RecommenderProtocol

to_dict()

Create an object’s dictionary representation.

Return type:

dict

Returns:

The dictionary representation of the object.

to_json(sink: str | Path | SupportsWrite[str] | None = None, /, *, overwrite: bool = False, **kwargs: Any)

Create an object’s JSON representation.

Parameters:

• sink (Union[str, Path, SupportsWrite[str], None]) –

  The JSON sink. Can be:

  • None (only returns the JSON string).

  • A file path or Path object pointing to a location where to write the JSON content.

  • A file-like object with a write() method.

• overwrite (bool) – Boolean flag indicating if to overwrite the file if it already exists. Only relevant if sink is a file path or Path object.

• **kwargs (Any) – Additional keyword arguments to pass to json.dumps().

Raises:

FileExistsError – If sink points to an already existing file but overwrite is False.

Return type:

str

Returns:

The JSON representation as a string.

property is_stateful: bool

Boolean indicating if the meta recommender is stateful.

mode: Literal['raise', 'reuse_last', 'cyclic']

Defines what shall happen when the last recommender in the sequence has been consumed but additional recommender changes are triggered:

• "raise": An error is raised.

• "reuse_last": The last recommender in the sequence is used indefinitely.

• "cycle": The selection restarts from the beginning of the sequence.

recommenders: tuple[RecommenderProtocol, ...]

A finite-length sequence of recommenders to be used. For infinite-length iterables, see baybe.recommenders.meta.sequential.StreamingSequentialMetaRecommender.