Probabilisty Sigmoid Output to Binary

Hi,

assume I have a tensor from a sigmoid output containing e.g. [0.1, 0.2, 0.3, 0.2, 0.4]. I now want [0, 0, 1, 1] i.e. for everything > 0.2 I want a 1 and else a 0.

How can I do that in pyTorch?

Hi
This can be done as follows =

output = model(input)
output = torch.sigmoid(output)
output = torch.round(output)

no, round has a fixed threshold of 0.5

Oh sorry I misread your question,
Then simply on your acitvated output do -

output = torch.where(output > 0.2, 1, 0)
1 Like