search

rail-berkeley / softlearning

Softlearning is a reinforcement learning framework for training maximum entropy policies in continuous domains. Includes the official implementation of the Soft Actor-Critic algorithm.
https://sites.google.com/view/sac-and-applications
Other
1.22k stars 242 forks source link

The issue of softlearning implementation #203

Open Ziyu0118 opened 1 year ago

Ziyu0118 commented 1 year ago

I attempted to implement softlearning with the usage mujoco210, but it appears to be unsuccessful. Is there currently an incompatibility issue between softlearning and mujoco210?

Ziyu0118 commented 1 year ago

The link to what I am trying is: http://proceedings.mlr.press/v139/xie21c

It is built upon the Tensorflow implementation of Soft Actor-Critic. The algorithm is experimented with four custom domains. I installed it entirely as the requirement.txt said, except some libraries are too old to use so I installed the oldest version currently available. So I would like to ask if anyone has successfully installed and used this article's environment before