rail-berkeley / rlkit

Collection of reinforcement learning algorithms
MIT License
2.5k stars 553 forks source link

Make sure you dont need a Mujoco license to use any of the algorithms #48

Closed redknightlois closed 5 years ago

redknightlois commented 5 years ago

There are many algorithms that import Mujoco environments because they are not separated. In my case I dont care about Mujoco, in fact I had to get a trial license just to avoid having to remove code from my fork.

vitchyr commented 5 years ago

I don't think any of the algorithm import mujoco. Do you have a specific file in mind? The example scripts certainly do. Is that what you're referring to?

redknightlois commented 5 years ago

In the chain of imports, for example VAE will import some environment proxies specific for mujoco. Is it ok if I move around that in a PR?

vitchyr commented 5 years ago

Oh I see. Yes, definitely feel free to make a PR for it.

gerritschoe commented 5 years ago

Outcommenting import mujoco_py in wrappers.py solved this for me. I am using Soft Actor-Critic and TD3 without Mujoco.

vitchyr commented 5 years ago

Closed with d3ea8168be40635317cbb1dc4ec9f292dee414a1