🐛 Bug

On master,

import torch
import torch.nn.functional as F
x = torch.randn(3, 3)
y = torch.randn(9)
F.logsigmoid(x, out=y) # crashes

The crash does not occur on CUDA on master.

However, in v1.4.0, the above snippet does not crash on CPU or CUDA.

Expected behavior

Either:

• No crash
• F.logsigmoid should not take an out argument. The docs imply there is no out argument: https://pytorch.org/docs/stable/nn.functional.html#torch.nn.functional.logsigmoid

Environment