You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
SB3 only logs the mean episodic return (i.e., rollout/ep_rew_mean), which is good enough in most cases. However, sometimes we may want the raw episodic return that is unaveraged. In my case, I was trying to compare my raw episodic return with rollout/ep_rew_mean, and it would look like something below, where SB3's orange curve would be much smoother because of the average operation.
Maybe it's worth also logging the raw stats, or at least give an option to.
Describe
SB3 only logs the mean episodic return (i.e.,
rollout/ep_rew_mean
), which is good enough in most cases. However, sometimes we may want the raw episodic return that is unaveraged. In my case, I was trying to compare my raw episodic return withrollout/ep_rew_mean
, and it would look like something below, where SB3's orange curve would be much smoother because of the average operation.Maybe it's worth also logging the raw stats, or at least give an option to.
#216 tracks the progress.
The text was updated successfully, but these errors were encountered: