def act(self, gs: GameState) -> int:
available_actions = gs.get_available_actions(gs.get_active_player())
if self.agents is None:
self.agents = [RandomAgent()] * gs.player_count()
accumulated_scores = np.zeros((len(available_actions),))
for i, a in enumerate(available_actions):
gs_clone = gs.clone()
gs_clone.step(gs.get_active_player(), a)
if self.determinist_environment:
max_scores = run_for_n_games_and_return_max(
self.agents, gs_clone, self.epochs_per_action
)
accumulated_scores[i] = max_scores[gs.get_active_player()]
else:
(total_scores, _, _) = run_for_n_games_and_return_stats(
self.agents, gs_clone, self.epochs_per_action
)
accumulated_scores[i] = total_scores[gs.get_active_player()]
# print((accumulated_scores, available_actions[np.argmax(accumulated_scores)]))
return available_actions[np.argmax(accumulated_scores)]
from drl_gym.agents import CommandLineAgent, RandomAgent
from drl_gym.environments.tictactoe import TicTacToeGameState
from drl_gym.runners import run_to_the_end
if __name__ == "__main__":
gs = TicTacToeGameState()
agent0 = CommandLineAgent()
agent1 = RandomAgent()
print(gs)
run_to_the_end([agent0, agent1], gs)
print(gs)
from drl_gym.agents import RandomAgent
from drl_gym.environments import GridWorldGameState
from drl_gym.runners import run_to_the_end
if __name__ == "__main__":
gs = GridWorldGameState()
agent = RandomAgent()
print(gs)
run_to_the_end([agent], gs)
print(gs)