The (originally combined) steps_per_second logging has been separated for now to enforce logging consistencies across Mava systems, enabling downstream aggregate plotting via marl-eval. However, the current steps_per_second logging in the IQL and SAC systems is not a true representation for the learner or the actor. It is not entirely fair to say that we do steps_per_rollout / elapsed _time steps per second in both cases because we are also taking anakin_steps * cfg.system.epochs gradient steps.
Describe the bug
The (originally combined)
steps_per_secondlogging has been separated for now to enforce logging consistencies across Mava systems, enabling downstream aggregate plotting viamarl-eval. However, the currentsteps_per_secondlogging in the IQL and SAC systems is not a true representation for the learner or the actor. It is not entirely fair to say that we dosteps_per_rollout / elapsed _timesteps per second in both cases because we are also takinganakin_steps * cfg.system.epochsgradient steps.