gryf/coach

mirror of https://github.com/gryf/coach.git synced 2025-12-17 11:10:20 +01:00

Files

guyk1971 74db141d5e SAC algorithm (#282 )

* SAC algorithm

* SAC - updates to agent (learn_from_batch), sac_head and sac_q_head to fix problem in gradient calculation. Now SAC agents is able to train.
gym_environment - fixing an error in access to gym.spaces

* Soft Actor Critic - code cleanup

* code cleanup

* V-head initialization fix

* SAC benchmarks

* SAC Documentation

* typo fix

* documentation fixes

* documentation and version update

* README typo

2019-05-01 18:37:49 +03:00

half_cheetah_sac.png

SAC algorithm (#282 )

2019-05-01 18:37:49 +03:00

hopper_sac.png

SAC algorithm (#282 )

2019-05-01 18:37:49 +03:00

humanoid_sac.png

SAC algorithm (#282 )

2019-05-01 18:37:49 +03:00

inverted_pendulum_sac.png

SAC algorithm (#282 )

2019-05-01 18:37:49 +03:00

README.md

SAC algorithm (#282 )

2019-05-01 18:37:49 +03:00

walker2d_sac.png

SAC algorithm (#282 )

2019-05-01 18:37:49 +03:00

README.md

Soft Actor Critic

Each experiment uses 3 seeds and is trained for 3M environment steps. The parameters used for SAC are the same parameters as described in the original paper.

Inverted Pendulum SAC - single worker

coach -p Mujoco_SAC -lvl inverted_pendulum

Hopper Clipped SAC - single worker

coach -p Mujoco_SAC -lvl hopper

Half Cheetah Clipped SAC - single worker

coach -p Mujoco_SAC -lvl half_cheetah

Walker 2D Clipped SAC - single worker

coach -p Mujoco_SAC -lvl walker2d

Humanoid Clipped SAC - single worker

coach -p Mujoco_SAC -lvl humanoid