- torch.flipud(input) Tensor ¶
Flip tensor in the up/down direction, returning a new tensor.
Flip the entries in each column in the up/down direction. Rows are preserved, but appear in a different order than before.
Requires the tensor to be at least 1-D.
torch.flipud makes a copy of
input’s data. This is different from NumPy’s np.flipud, which returns a view in constant time. Since copying a tensor’s data is more work than viewing that data, torch.flipud is expected to be slower than np.flipud.
input (Tensor) – Must be at least 1-dimensional.
>>> x = torch.arange(4).view(2, 2) >>> x tensor([[0, 1], [2, 3]]) >>> torch.flipud(x) tensor([[2, 3], [0, 1]])