PlaidSearcher

class lightning_ir.retrieve.plaid.plaid_searcher.PlaidSearcher(index_dir: Path | str, search_config: PlaidSearchConfig, module: BiEncoderModule, use_gpu: bool = False)

Bases: Searcher

Searcher for Plaid, a residual-based search method for efficient retrieval.

__init__(index_dir: Path | str, search_config: PlaidSearchConfig, module: BiEncoderModule, use_gpu: bool = False) → None

Initialize the PlaidSearcher.

Parameters:

• index_dir (Path | str) – Directory where the Plaid index is stored.

• search_config (PlaidSearchConfig) – Configuration for the Plaid searcher.

• module (BiEncoderModule) – The BiEncoder module used for searching.

• use_gpu (bool) – Whether to use GPU for searching. Defaults to False.

Methods

__init__(index_dir, search_config, module[, ...])	Initialize the PlaidSearcher.
search(output)	Search for relevant documents using the Plaid index.

search(output: BiEncoderOutput) → Tuple[PackedTensor, List[List[str]]]

Search for relevant documents using the Plaid index.

Parameters:

output (BiEncoderOutput) – The output from the BiEncoder module containing query embeddings.

Returns:

A tuple containing the scores and the corresponding document IDs.

Return type:

Tuple[PackedTensor, List[List[str]]]

Raises:

ValueError – If the output does not contain query embeddings.

to_gpu() → None

Move the searcher to the GPU if available.