"""Base class of all dataset iterators.
Iterator iterates over the dataset, yielding a minibatch at each
iteration. Minibatch is a list of examples. Each implementation should
implement an iterator protocol (e.g., the :meth:`__next__` method).
Note that, even if the iterator supports setting the batch size, it does
not guarantee that each batch always contains the same number of examples.
For example, if you let the iterator to stop at the end of the sweep, the
last batch may contain a fewer number of examples.
The interface between the iterator and the underlying dataset is not fixed,
and up to the implementation.
Each implementation should provide the following attributes (not needed to
- ``batch_size``: the number of examples within each minibatch.
- ``epoch``: the number of completed sweeps over the dataset.
- ``epoch_detail``: floating point number version of the epoch. For
example, if the iterator is at the middle of the dataset at the third
epoch, then this value is 2.5.
- ``is_new_epoch``: True if the epoch count was incremented at the last
Each implementation should also support serialization to resume/suspend the
[docs] def __iter__(self):
[docs] def __next__(self):
"""Returns the next batch.
This is a part of the iterator protocol of Python. It may raise the
:class:`StopIteration` exception when it stops the iteration.
[docs] def next(self):
"""Python2 alternative of ``__next__``.
It calls :meth:`__next__` by default.
[docs] def finalize(self):
"""Finalizes the iterator and possibly releases the resources.
This method does nothing by default. Implementation may override it to
better handle the internal resources.
[docs] def serialize(self, serializer):
"""Serializes the internal state of the iterator.
This is a method to support serializer protocol of Chainer.
It should only serialize the internal state that changes over the
iteration. It should not serializes what is set manually by
users such as the batch size.