[bug-fix] Improve performance for PPO with continuous actions#3662

Merged

ervteng merged 4 commits intomasterfrom

develop-gaussiandistribution

Mar 20, 2020

Commits on Mar 19, 2020

Fix bug with tanh + reparameterize
Ervin Teng
committed
Sigma in gaussian distribution to be unconditioned
Ervin Teng
committed

Commits on Mar 20, 2020

Lower success threshold for recurrent PPO
Ervin Teng
committed
More stable PPO recurrent test
Ervin Teng
committed