Fixes "turn" game state tracking in inference and add n_steps command-line arg

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

Fixes "turn" game state tracking in inference and add n_steps command-line arg #55

Closed glmcdona closed 3 years ago

glmcdona commented 3 years ago

Fixes an important bug where state["turn"] was not being populated in inference. This means that agents in deployment would perform more poorly if they relied on this feature as part of their observation space.