Improves the RecordEpisodeMetrics wrapper by adding dones, which allows us to pull out the final reward from the episode. Thus instead of logging the mean of many final episode returns we now log the mean of just the last episode return.
Why?
More accurate representation of system performance.
What?
Improves the
RecordEpisodeMetricswrapper by adding dones, which allows us to pull out the final reward from the episode. Thus instead of logging the mean of many final episode returns we now log the mean of just the last episode return.Why?
More accurate representation of system performance.