a392ddaeae3a8343a968c8839afeea70b6f3a8b8,thinc/layers/recurrent.py,,recurrent,#,6

Before Change

def recurrent(step_model: Model) -> Model:
    return Model(step_model.name, forward, layers=[step_model])


def forward(model: Model, X_size_at_t: Tuple[Array, Array], is_train: bool):
    // Expect padded batches, sorted by decreasing length. The size_at_t array

After Change

        dims={"nO": step_model.get_dim("nO") if step_model.has_dim("nO") else None},
        layers=[step_model]
    )
    if model.has_dim("nO"):
        model.initialize()
    return model


def init(model, X=None, Y=None):
    Xt = X[0][0] if X is not None else None

Instances