Skip to content

CyCTW/SurakartaAI

 
 

Repository files navigation

SurakartaAI

This Repository build a Surakarta Game framework and implement MCTS (Monte Carlo Tree Search) technique to train AI. In the future, we will combine MCTS with Deep Q-Learning or other appropriate RL, ML, DRL technique.

How to Run?

Important: You should install C++ libtorch libaray first

  1. git clone this Repository
  2. cd SurakartaAI
  3. mkdir build
  4. cd build
  5. cmake -DCMAKE_PREFIX_PATH=/path/to/yourlibtorch ..
  6. cmake --build .
  7. ./surakarta (--total=1000 --block=100 (default argument))

How to Modify Strategy?

  • By deault, The two agent player and envir move according to two different strategy, MCTS and Eat First Greedy.
  • Modify agent.h and choose the strategy you want in policy.h

Releases

No releases published

Packages

No packages published

Languages

  • C++ 99.3%
  • CMake 0.7%