Dataset of tuples from multiple equal-length datasets.
TupleDatasetcombines multiple equal-length datasets into a single dataset of tuples. The
i-th tuple contains the
i-th example from each of the argument datasets, in the same order that they were supplied.
Recall that in Chainer, a dataset is defined as an iterable that supports both
__getitem__method should support indexing by both an integer and a slice.
As an example, consider creating a
TupleDatasetfrom two argument datasets
d1 = [8, 0, 5, 1]and
d2 = [3, 1, 7, 4]as
tuple_dataset = TupleDataset(d1, d2). The
tuple_datasetwill then contain the examples
(8, 3), (0, 1), (5, 7), (1, 4). Note that this behavior is similar to that of the built-in
datasets – Underlying datasets that will be aggregated. Each dataset must be an iterable that implements
j-th dataset will be used for the
j-th item of each example tuple. All datasets must have the same length.