ryanjulian / rllab

rllab is a framework for developing and evaluating reinforcement learning algorithms, fully compatible with OpenAI Gym.
Other
1 stars 21 forks source link

Refactor rllab from rllab.mujoco_py to mujoco_py (v1.5.0) #91

Closed jonashen closed 6 years ago

jonashen commented 6 years ago

When rllab was first developed, the mujoco_py package written by MuJoCo was barebones and sorely lacking in functionality. As such, the original creators of rllab decided to copy in a local folder called rllab.mujoco_py, in which they built upon the mujoco_py package with custom functions they felt were necessary to implement rllab.

As MuJoCo has now upgraded to a new version (i.e. v1.5.0), many of the functions in this folder are now obsolete, and can be replaced. Thus, there is no need to keep this folder; instead, we can utilize the functionality of mujoco_py directly. This PR will replace all usage of rllab.mujoco_py with mujoco_py instead, as well as remove the rllab.mujoco_py folder itself.

P.S. this PR will also update the setup_mujoco.sh script to support installation of MuJoCo v1.5.0 in lieu of v1.3.0.

P.P.S. I have also inadvertently built upon a commit that supports asynchronous support for plot rollout.

See: #1, #4, #55, #90