def move_down(self, matrix):
if matrix in self.down_table:
self.hits += 1
return self.down_table[matrix]
else:
self.down_table[matrix] = game.move_down(matrix)
return self.down_table[matrix]
def update_frame(x):
global state, score, high_score, last_move, bot_mode, down_press
if bot_mode:
# a = policy(Variable(torch.from_numpy(state).type(torch.FloatTensor)))
# _, ac = a.max(0)
# action = ac.item()
a = policy(Variable(torch.from_numpy(state).type(torch.FloatTensor)))
# a = F.softmax(a, dim=-1)
c = Categorical(a)
action = c.sample()
# action = train.select_action(state).item()
state, reward, done = game.step(action)
last_move = action
else:
state, reward, done = game.step(0)
if down_press:
game.active_piece, game.grid, _ = game.move_down(game.active_piece, game.grid)
last_move = 4
score += reward
if done:
game.reset()
high_score = max(high_score, score)
score = 0
def move_down(matrix):
if matrix in DOWN_TABLE:
global HITS
HITS += 1
return DOWN_TABLE[matrix]
else:
DOWN_TABLE[matrix] = game.move_down(matrix)
return DOWN_TABLE[matrix]
def action(x,i):
if i == 0:
return game.move_up(x)
elif i == 1:
return game.move_down(x)
elif i == 2:
return game.move_left(x)
elif i == 3:
return game.move_right(x)
def action(loc, board):
if loc == 0:
board = game.move_up(board)
elif loc == 1:
board = game.move_down(board)
elif loc == 2:
board = game.move_left(board)
else:
board = game.move_right(board)
return board
def find_best_move(matrix):
best_move = None
best_score = -1
matrixes = [game.move_up(matrix),
game.move_down(matrix),
game.move_left(matrix),
game.move_right(matrix)]
for i in range(4):
if matrixes[i] != matrix:
score = score_matrix(matrixes[i])
if score >= best_score:
best_score = score
best_move = i
return best_move, best_score
def action(x, board):
comp = 0
loc = 0
for i in range(4):
if x[i] > comp:
loc = i
comp = x[i]
if loc == 0:
board = game.move_up(board)
elif loc == 1:
board = game.move_down(board)
elif loc == 2:
board = game.move_left(board)
else:
board = game.move_right(board)
return board