Skip to content

Asynchronous server for collecting offline rollouts in a reinforcement learning setting

Notifications You must be signed in to change notification settings

mdkozlowski/surge

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

22 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

surge

Asynchronous server for collecting offline rollouts in a reinforcement learning setting

  1. Externally, Pytorch models of agent policy functions are are trained using PPO
  2. Models weights are are sent by clients to be cached in the server
  3. Each model version plays multiple matches against all other models
  4. Rollouts of these matches are collected and returned to the clients

A fruitbots clone is used as the game environment in this engine

About

Asynchronous server for collecting offline rollouts in a reinforcement learning setting

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages