Wrap x and y in a DataLoader with shuffle=False and batch_size, then return the list of (feature_batch, label_batch) pairs it yields.

With shuffling off the order is the dataset's own, and the final batch is whatever is left over -- it is not padded or dropped unless you ask for that.

Input

x =
tensor([[0., 1.],
        [2., 3.],
        [4., 5.],
        [6., 7.],
        [8., 9.]])
y = tensor([0, 1, 2, 3, 4])
batch_size = 2

Output

[(tensor([[0., 1.],
        [2., 3.]]), tensor([0, 1])), (tensor([[4., 5.],
        [6., 7.]]), tensor([2, 3])), (tensor([[8., 9.]]), tensor([4]))]