def choose_action(self, state, greedy=False):
# 决策
if greedy or np.random.random() < self.epsilon:
board = np.expand_dims(state, axis=0)
actions_value = self.model1.predict(board)
action = np.argmax(actions_value)
print('=====================================')
print('action_value', actions_value)
print('-------------------------------------')
print(action)
print('=====================================')
else:
# action = np.random.randint(0, 4)
gameboard = np.zeros((4, 4))
for i in range(4):
for j in range(4):
num = np.argmax(state[i][j])
if num == 0:
gameboard[i][j] = 0
else:
gameboard[i][j] = 2**num
action = board_to_move(gameboard)
return action
def move(self, game):
ohe_board = grid_one(vmap(game.board))
d = board_to_move(game.board)
self.mem.push(ohe_board, d)
if random.random() < 0.6:
game.move(d)
else:
game.move(self.predict(ohe_board).argmax())
def move(self, game):
ohe_board = game.board
suggest = board_to_move(game.board)
direction = self.predict(game.board).argmax()
game.move(direction)
a = random.randint(0, 9)
self.memory.push(ohe_board, suggest)
return game
def move(self, game):
ohe_board = grid_ohe(game.board)
from game2048.expectimax import board_to_move
suggest = board_to_move(game.board)
direction = self.predict(ohe_board).argmax()
game.move(direction)
self.memory.push(ohe_board, suggest)
def move(self, game):
ohe_board = grid_one(vmap(game.board))
d = board_to_move(game.board)
self.mem.push(ohe_board, d)
game.move(self.predict(ohe_board).argmax())
def move(self, game):
ohe_board = grid_one(vmap(game.board))
suggestion = board_to_move(game.board)
direction = self.predict(ohe_board).argmax()
game.move(direction)
self.mem.push(ohe_board, suggestion)
for i3 in range(8):
gameeasy = Game(4, score_to_win=256, random=False)
agente = ExpectiMaxAgent(gameeasy)
agente.play(verbose=True)
'''
#***********************************************************************
#*********************************************************************
#boardsets=np.loadtxt('b1024-'+str(i+1)+'.csv',usecols=(0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15),delimiter=',')
#boardsets=csv.reader(open('resulttest.csv'))
#for row in boardsets:
#print row
boardsets = yboard
rest = np.empty(shape=[100000, 17])
i3 = 0
print(1)
for item2 in boardsets:
tmpitem2 = item2[0:16]
boardtmp = tmpitem2.reshape(4, 4)
direction = board_to_move(boardtmp)
item2tmp = np.append(tmpitem2, [direction], axis=0)
rest[i3] = item2tmp
i3 = i3 + 1
saveboard = rest[0:i3]
#with open('b2048-'+str(i+1)+'.csv', 'a', newline='') as csv_file:
#csv_writer = csv.writer(csv_file)
#for iw in range(i3):
#csv_writer.writerow(rest[iw])
def play(self):
onehot_board = change_to_onehot(self.game.board)
self.push(onehot_board, board_to_move(self.game.board))
self.game.move(self.predict())