kaggle-Lux-AI

BigNoBrainer

• Perception class feeds the current game state to the network
• A single large network translates the perception into (?)
• Executor class follows the hints from the (?) output matricies and emits unit and citytile actions.

A module construct a gif of the input matricies from a Game() class. The gif below is made by compiling the Game() states every 50 turns of the rule based clusterbot L16H7

test_json_processsor.py

This python script converst from a replay.json that can be downloaded from Kaggle in JSON format to a Game() class, to a Perception() class and finally to a .gif image.
It takes about 90m to generate the .gif Kaggle Replay episode 27883823