hongshaorou/Tianshou
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
Tianshou/examples/box2d
History
ChenDRAG 7036073649
Trainer refactor : some definition change (#293) 
This PR focus on some definition change of trainer to make it more friendly to use and be consistent with typical usage in research papers, typically change `collect-per-step` to `step-per-collect`, add `update-per-step` / `episode-per-collect` accordingly, and modify the documentation.
2021-02-21 13:06:02 +08:00
..
results/sac
sac mujoco result (#246)
2020-11-09 16:43:55 +08:00
acrobot_dualdqn.py
Trainer refactor : some definition change (#293)
2021-02-21 13:06:02 +08:00
bipedal_hardcore_sac.py
Trainer refactor : some definition change (#293)
2021-02-21 13:06:02 +08:00
lunarlander_dqn.py
Trainer refactor : some definition change (#293)
2021-02-21 13:06:02 +08:00
mcc_sac.py
Trainer refactor : some definition change (#293)
2021-02-21 13:06:02 +08:00
README.md
sac mujoco result (#246)
2020-11-09 16:43:55 +08:00

README.md

Bipedal-Hardcore-SAC

  • Our default choice: remove the done flag penalty, will soon converge to ~280 reward within 100 epochs (10M env steps, 3~4 hours, see the image below)
  • If the done penalty is not removed, it converges much slower than before, about 200 epochs (20M env steps) to reach the same performance (~200 reward)