A wrapper that unrolls the second (time) dimension of a tensor into the first (batch) dimension, applies some other Module, and then rolls the time dimension back up.

class allennlp.modules.time_distributed.TimeDistributed(module)[source]

Bases: torch.nn.modules.module.Module

Given an input shaped like (batch_size, time_steps, [rest]) and a Module that takes inputs like (batch_size, [rest]), TimeDistributed reshapes the input to be (batch_size * time_steps, [rest]), applies the contained Module, then reshapes it back.

Note that while the above gives shapes with batch_size first, this Module also works if batch_size is second - we always just combine the first two dimensions, then split them.

It also reshapes keyword arguments unless they are not tensors or their name is specified in the optional pass_through iterable.

forward(*inputs, pass_through: typing.List[str] = None, **kwargs)[source]

Defines the computation performed at every call.

Should be overridden by all subclasses.


Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.