issues
search
glmcdona
/
LuxPythonEnvGym
Matching python environment code for Lux AI 2021 Kaggle competition, and a gym interface for RL models.
MIT License
73
stars
38
forks
source link
Simplify example training script and agent
#87
Closed
glmcdona
closed
2 years ago
glmcdona
commented
2 years ago
Simplify reward, metrics, and logic.
Reward is now just a delta reward per unit, citytile, and resources collected. Also a reward for each city tile end of game.
Removed stats logging.
Removed more smart city research logic.
Tuned some comments.
Removed train script game-details tensorflow custom logger.
Simplify reward, metrics, and logic.