Builder Trains Cartpole In MuJoCo At 18 Million Steps Per Second
--0--
Sentiment
Sentiment building, check back later.
Cluster Engagement
-
Views
-
Comments
-
Reposts
-
Bookmarks
Expand data
Posts from X
Most Activity
Most ActivityTimeline
VIEWS41KBOOKMARKS41LIKES191REPLIES19
kache@yacineMTB
i love this stuff so much
kache@yacineMTB
I just trained cartpole in mujoco at 18 million steps per second. This policy learned in **less than 3 seconds**
rollout policy batch size was 8192 agents
12hViews 41KLikes 191Bookmarks 41
RETWEETS5
kache@yacineMTB
apparently quadruple inverted pendulums were only solved a year ago..!
so I guess if someone solves 5 pendulums, they'll be the first in the world ever. i mean how hard could it possibly be?
kache@yacineMTB
i wonder when this task becomes impossible to solve
12hViews 14.6KLikes 111Bookmarks 23