espnet.nets.beam_search_partially_AR.PartiallyARHypothesis
Less than 1 minute
espnet.nets.beam_search_partially_AR.PartiallyARHypothesis
class espnet.nets.beam_search_partially_AR.PartiallyARHypothesis(yseq: Tensor, score: float | Tensor | None = None, states: Dict[str, Any] = {}, yseq_length: Tensor = tensor([]), eos: Tensor = tensor([]))
Bases: NamedTuple
Hypothesis data type for partially autoregressive decoding.
Create new instance of PartiallyARHypothesis(yseq, score, states, yseq_length, eos)
asdict() → dict
Convert data to JSON-friendly dict.
eos : Tensor
Alias for field number 4
score : float | Tensor
Alias for field number 1
states : Dict[str, Any]
Alias for field number 2
yseq : Tensor
Alias for field number 0
yseq_length : Tensor
Alias for field number 3