vertebrae.extractors.precomputed

Extractor for precomputed embeddings.

Classes

PrecomputedExtractor

Extractor for already-computed dense or sparse embeddings.

Module Contents

class vertebrae.extractors.precomputed.PrecomputedExtractor(name='precomputed', *, cache_embeddings=True)[source]

Extractor for already-computed dense or sparse embeddings.

Parameters:

  • name (str) – User-facing extractor name.

  • cache_embeddings (bool) – Whether benchmark workflows may cache the supplied embeddings.

fit(X, y=None)[source]

No-op fit for precomputed embeddings.

Parameters:

  • X (Any) – Embedding matrix.

  • y (Any) – Optional labels.

Returns:

This extractor.

Return type:

PrecomputedExtractor

transform(X)[source]

Validate and return precomputed embeddings.

Parameters:

X (Any) – Dense or sparse embedding matrix.

Returns:

Validated dense or sparse numeric embedding matrix.

Return type:

numpy.ndarray

fit_transform(X, y=None)[source]

Return validated precomputed embeddings.

Parameters:

  • X (Any) – Dense or sparse embedding matrix.

  • y (Any) – Optional labels.

Returns:

Validated dense or sparse numeric embedding matrix.

Return type:

numpy.ndarray

recipe()[source]

Return a serializable recipe for this extractor.

Returns:

JSON-compatible recipe dictionary.

Return type:

Dict[str, Any]