I create a 2048 game solver with Unity machine learning agent. But it can only read 256 tile. I really wish any one can help me improve it.
- I train in python for 30 minutes with 50k step.
- My state: 17 member: 1 member is size of board, and 16 member is value at every tile.
- Action space is 4 and type of action space is discrete. It return a direction to move the board.
- Github: https://github.com/pearl2201/Unity-2048-ML