Hello, In this paper, SMAC with sparse rewards needs many time steps (0.2*10^8) to converge. I want to know how much time step does it need when I want to train SMAC with dense rewards. ( I am considering '3m')