Basic DeepQ in pytorch, reworked scoring systems to allow multiple di…

harman097 / RoboRugby

Training environment for ML agents in a competitive robot rugby game using OpenAI's gym standard.

MIT License

4 stars 0 forks source link

Closed harman097 closed 4 years ago

harman097 commented 4 years ago

Added new "GameEnv_Simple" with discrete action space + smaller observation space
Reworked scoring systems as new RR_Scorekeepers classes to allow multiple different combos of scoring to co-exist and easily swap in/out without touching base GameEnv
Added basic DQN implementation + sample training loop using Pytorch