HigashikataZhangsuke/2048-api

A 2048 api for training supervised learning (imitation learning) or reinforcement learning agents

2048-api

A 2048 game api for training supervised learning (imitation learning) agents

Requirements

• code only tested on linux system (ubuntu 16.04)
• Python 3 (Anaconda 3.6.3 specifically) with numpy and flask

#代码结构：

• game2048/: the main package.
  • game.py: the core 2048 Game class.
  • agents.py: the Agent class with instances.I added TrainAgent and TestAgent classes.
  • displays.py: the Display class with instances, to show the Game state.
  • expectimax/: a powerful ExpectiMax agent by here.
  • CNN.py: a convolution nueral network for trainning.
  • one_hotboard.py: transfer board to onehot board.
  • calculatedata.py: use some properties of the board let train faster.
• explore.ipynb: introduce how to use the Agent, Display and Game.
• static/: frontend assets (based on Vue.js) for web app.
• webapp.py: run the web app (backend) demo.
• evaluate.py: evaluate your self-defined agent.
• 'train.py': used to train CNN

#代码运行：将git网址下的代码clone到本地后在pycharm中新建project并打开，直接运行evaluate，等到控制台输出结束命令时在my2048-api文件夹下的EE369.log文件中可以得到将结果。或者在my2048-api中打开终端，输入chmod a+x evaluate.py，回车后再输入python evaluate.py即可。权重文件是model2048.pkl。