hongshaorou/Tianshou
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
Tianshou/examples/box2d
History
n+e 5ed6c1c7aa
change the step in trainer (#235) 
This PR separates the `global_step` into `env_step` and `gradient_step`. In the future, the data from the collecting state will be stored under `env_step`, and the data from the updating state will be stored under `gradient_step`.

Others:
- add `rew_std` and `best_result` into the monitor
- fix network unbounded in `test/continuous/test_sac_with_il.py` and `examples/box2d/bipedal_hardcore_sac.py`
- change the dependency of ray to 1.0.0 since ray-project/ray#10134 has been resolved
2020-10-04 21:55:43 +08:00
..
results/sac
SAC implementation update (#212)
2020-09-12 08:44:50 +08:00
acrobot_dualdqn.py
change API of train_fn and test_fn (#229)
2020-09-26 16:35:37 +08:00
bipedal_hardcore_sac.py
change the step in trainer (#235)
2020-10-04 21:55:43 +08:00
lunarlander_dqn.py
change API of train_fn and test_fn (#229)
2020-09-26 16:35:37 +08:00
mcc_sac.py
change API of train_fn and test_fn (#229)
2020-09-26 16:35:37 +08:00
README.md
SAC implementation update (#212)
2020-09-12 08:44:50 +08:00

README.md

Bipedal-Hardcore-SAC

  • Our default choice: remove the done flag penalty, will soon converge to ~270 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)