Does torch.jit.script need to specify device?

For using torch.jit.script is recommended to pass model to cpu if you are doing inference in that device?