Similarities of Sentence Pairs

Our package also provides a way to easily calculate (cosine) similarities between two sentences:

model.similarity(queries, keys, device=None)

Inputs

• queries: a string or a list of n strings.
• keys: a string or a list of m strings.
• device: cuda or cpu.

Outputs

• A float if queries and keys are both single sentence. Otherwise the function returns a numpy array of size n * m.