vwxyzjn / vectorized-value-methods

[WIP] Vectorized architecture for value-based methods such as DQN and DDPG
https://docs.google.com/document/d/1hdRAXtqNunmcyvULkCnDOg56UfnOFYq2KqX9lSRVfyM/edit
MIT License
3 stars 2 forks source link

Port over soft actor critic #8

Closed vwxyzjn closed 2 years ago

vwxyzjn commented 2 years ago

Just setting up an issue to track this progress. @lockwo had previously expressed interest in this. A related resource is CleanRL is now introducing a refactored sac implementation in https://github.com/vwxyzjn/cleanrl/pull/102 (thanks to @dosssman), which has good performance and might make it easier to port SAC over here.

vwxyzjn commented 2 years ago

Addressed in #10.