update SAC Roboschool and full Humanoids benchmark - Githubissues

kengz / SLM-Lab

Modular Deep Reinforcement Learning framework in PyTorch. Companion library of the book "Foundations of Deep Reinforcement Learning".

https://slm-lab.gitbook.io/slm-lab/

MIT License

1.25k stars 264 forks source link

update SAC Roboschool and full Humanoids benchmark #409

Closed kengz closed 5 years ago

kengz commented 5 years ago

Feature / Fix

update SAC benchmark from rerun

Click on the algorithm to see the result upload Pull Request.

Env. \ Alg.	A2C (GAE)	A2C (n-step)	PPO	SAC
RoboschoolAnt graph	1029.51 graph	1148.76 graph	1931.35 graph	2914.75 graph
RoboschoolAtlasForwardWalk graph	68.15 graph	73.46 graph	148.81 graph	942.39 graph
RoboschoolHalfCheetah graph	895.24 graph	409.59 graph	1838.69 graph	2496.54 graph
RoboschoolHopper graph	286.67 graph	-187.91 graph	2079.22 graph	2251.36 graph
RoboschoolInvertedDoublePendulum graph	1769.74 graph	486.76 graph	7967.03 graph	8085.04 graph
RoboschoolInvertedPendulum graph	1000.0 graph	997.54 graph	930.29 graph	941.45 graph
RoboschoolReacher graph	14.57 graph	-6.18 graph	19.18 graph	19.99 graph
RoboschoolWalker2d graph	413.26 graph	141.83 graph	1368.25 graph	1894.05 graph

Humanoid environments are significantly harder. Note that due to the number of frames required, we could only run Async-SAC.

Env. \ Alg.	A2C (GAE)	A2C (n-step)	PPO	Async-SAC
RoboschoolHumanoid	122.23 graph	-6029.02 graph	1554.03 graph	2621.46 graph
RoboschoolHumanoidFlagrun	93.48 graph	-2079.02 graph	1635.64 graph	1937.77 graph
RoboschoolHumanoidFlagrunHarder	-472.34 graph	-24620.71 graph	610.09 graph	280.18 graph