Open RL Benchmark by CleanRL

The code is available at https://github.com/vwxyzjn/cleanrl, and this file benchmark.sh produced the runs that are used in this report.

In this page, we showed the performances of various DRL algorithms in diverse game environments. It gives us important information regarding the learnability and generality of algorithms. The charts are interactive. Feel free to add different visualization and click on a specific run to see its hyper-parameters.

CartPole-v0

MountainCar-v0

MountainCarContinuous-v0

Walker2DBulletEnv-v0

HumanoidBulletEnv-v0

ReacherBulletEnv-v0

BipedalWalker-v2

Pendulum-v0

HopperBulletEnv-v0

InvertedPendulumBulletEnv-v0

LunarLander-v2

Taxi-v3