Factored Diffusion Policies:Compositionally Generalized Robot Control with a Single Score Network | Steady Practice | SteadyPractice