Hello.
I have question.
Could I change the weight and bias in nn.torch.Sigmoid()?
Hello.
I have question.
Could I change the weight and bias in nn.torch.Sigmoid()?
The sigmoid function does not hold any weight or bias parameters. It just applies 1 / (1 + e**(-x))
on its input x
.
Thanks for your reply.
Then, If I use Sigmoid function, Couldn’t revise bias and weight??
I’m not sure I understand your use case clearly.
If you would like to take a look at some parameters, you can call them directly in the corresponding module:
weight = model.fc.weight
bias = model.fc.bias
where your model contains e.g. a linear layer assigned to self.fc
.