Yes I have noticed that. So in my latest code I changed it to
c = pyro.param("c", torch.tensor(0.001), constraint=constraints.positive)
(see the update in my original post). But there is still error. By the way I have also tried
c = pyro.param("c", torch.tensor(0.001), constraint=constraints.greater_than_eq(0))
and it didn’t work either.