def do_rollout(self, node: GameState):
node = node.__copy__()
"Make the tree one layer better. (Train for one iteration.)"
path = self._select(node)
leaf = path[-1]
self._expand(leaf)
if self.heuristic is None:
reward = self._simulate(leaf.__copy__())
else:
reward = self.heuristic.h(node)
self._backpropagate(path, reward)
def play(self, game: GameState, verbose=False):
if self.alternate:
temp = self.actionModel1
self.actionModel1 = self.actionModel2
self.actionModel2 = temp
states = []
hs = []
if verbose:
print(game)
while not game.game_over():
actingModel = self.actionModel1 if game.turn() else self.actionModel2
move, h = actingModel.action(game)
states.append(game.__copy__())
hs.append(h)
game.push(move)
if verbose:
print("move: " + str(move) + " h: " + str(h))
print(game)
return game.winner(), states, hs
