allennlp.models.ensemble

class allennlp.models.ensemble.Ensemble(submodels: typing.List[allennlp.models.model.Model]) → None[source]

Bases: allennlp.models.model.Model

An ensemble runs multiple instances of a model and selects an answer from the subresults via some ensembling strategy.

Ensembles differ from most models in that they do not have a vocabulary or weights of their own (instead they rely on the vocabulary and weights from submodels). Instead, the submodels are trained independently and the ensemble is created from the result.