seqme.Cache

class seqme.Cache(models=None, init_cache=None)

Caches model-generated feature representations of sequences.

Allows storing and retrieving representations per model to avoid recomputation, with support for adding models and precomputed values.

__init__(models=None, init_cache=None)

Initialize the cache with optional models and precomputed representations.

Parameters:

• models (Optional[dict[str, Callable[[list[str]], list[Any] | ndarray]]]) – Mapping from model name to callable for generating representations.

• init_cache (Optional[dict[str, dict[str, ndarray]]]) – Initial cache of sequence feature representations by model and sequence.

__call__(sequences, model_name, stack)

Return feature representations for the given sequences using the specified model.

Uncached sequences are computed and stored.

Parameters:

• sequences (list[str]) – List of text sequences.

• model_name (str) – Name of the model to use.

• stack (bool) – Whether the feature representations should be stacked as a numpy array. If True then stack as a numpy array else return a list of representations.

Return type:

list[Any] | ndarray

Returns:

Feature representations in the same order as the input sequences.

Methods

__init__([models, init_cache])	Initialize the cache with optional models and precomputed representations.
__call__(sequences, model_name, stack)	Return feature representations for the given sequences using the specified model.
add(model_name, element)	Add a new model or precomputed representations to the cache.
get()	Return a copy of the current cache.
model(model_name, *[, stack])	Return a callable interface for a given model name.
remove(model_name)	Remove the cache of a model and the model callable if defined.