Stumbled across something odd as I was playing around with tensor slicing.
a = torch.arange(20).view(4, 5)
# tensor([[ 0., 1., 2., 3., 4.],
# [ 5., 6., 7., 8., 9.],
# [ 10., 11., 12., 13., 14.],
# [ 15., 16., 17., 18., 19.]])
If I use invalid indices for the first dimension (e.g., a[5:, :], a[4:4, :]), an error pops up (as expected).
# Traceback (most recent call last):
# File "<stdin>", line 1, in <module>
# RuntimeError: dimension out of range (expected to be in range of [-1, 0], but got 1)
But when I use them for the second dimension, this happens:
a[:, 5:]
# tensor([ 5., 10., 15., 0.])
# Permutation of the first column?
a[:, 4:4]
# tensor([ 4., 9., 14., 19.])
# Squeezed final column
Is this intentional/does this serve a purpose? Thanks in advance!