How to optimize parameters in a specified range in PyTorch?

Question

A trainable parameter p initialized to 0.5:

class Model(nn.Module):
    def __init__(self):
        super(Model, self).__init__()
        self.p = nn