Some confusion about saving model parameters

I met a problem, in my model, one layer is like:

class Model():
     def __init__():
         self.w = nn.Linear(1,2)
         self.b = Variable(torch.randn(2))
     def forward(x):
         out = self.w(x) + self.b

I know there is bias on Linear model, but I want to get another bias, but after, the self.b can not be saved.
Hence, what is the right way to save the model so that I can use the self.b from saved model?