data

Data Iterators

Designed to feed the network data for training.

Iterators return Batch datatype, which is named tuple with inputs and targets tensors:

Batch = NamedTuple("Batch", [("inputs", Tensor), ("targets", Tensor)])

Classes

Batch(inputs, targets)
BatchIterator(batch_size, shuffle)	Default iterator, returns batches of input, target pairs.
DataIterator()	Base class that every data iterator inherits.