Use Sigmoid Top prediction model but i didn't the probabality always output 0.0

vision
1

hey to all of you
i would to like to know why. when i am using torch.sigmoid(Y_hat) to convert to probability, always get 0.0 without extra values

2

torch.sigmoid will output zeros for large negative values as seen here:

torch.sigmoid(torch.ones(10) * -1000)
# tensor([0., 0., 0., 0., 0., 0., 0., 0., 0., 0.])

I don’t know which “extra values” you are expecting so could you describe your use case and issue a bit more, please?