sparank.data.InferenceDataset

class sparank.data.InferenceDataset(tokens, context_ids=None)

Bases: Dataset

Minimal in-memory dataset designed for inference.

Parameters:

• tokens (np.ndarray) – Integer token matrix produced by tokenize_batch, shape (N, seq_len).

• context_ids (np.ndarray, optional) – Optional context / state IDs per sample, shape (N,).

__init__(tokens, context_ids=None)

Parameters:

• tokens (ndarray)

• context_ids (ndarray | None)

Methods

__init__(tokens[, context_ids])