class SnakeGameDemoWidget(QtWidgets.QWidget):
"""
SnakeGame
"""
def __init__(self):
super(SnakeGameDemoWidget, self).__init__()
self.resize(QtCore.QSize(300, 300))
self.setWindowTitle("SnakeGame")
self.snake_game = SnakeGame()
self.snake_game.deterministic_food = True
self.snake_game.food_positions = [
(6, 6),
(2, 15),
(17, 3)
]
self.snake_game.rect = self.rect()
self.snake_game.width = self.width()
self.snake_game.height = self.height()
self.snake_game.setup()
self.tick_timer = QTimer()
self.tick_timer.setInterval(100)
self.tick_timer.timeout.connect(self.tick)
def showEvent(self, event):
self.snake_game.rect = self.rect()
self.tick_timer.start()
def resizeEvent(self, event):
self.snake_game.rect = self.rect()
def closeEvent(self, event):
self.tick_timer.stop()
def tick(self):
self.snake_game.tick()
self.update()
def paintEvent(self, event):
painter = QtGui.QPainter(self)
painter.setRenderHint(QtGui.QPainter.Antialiasing, True)
self.snake_game.paint(painter)
def keyPressEvent(self, event):
self.snake_game.key_pressed(event.key())
def sizeHint(self):
return QtCore.QSize(300, 300)
def train(num_episodes):
max_moves_without_fruit = 15
for i in range(num_episodes):
game = SnakeGame(width, height, num_fruit=3)
num_moves_without_fruit = 0
while not game.game_over:
observation = numpy.copy(game.board)
action = choose_action(snake_model, observation)
game.tick(game_actions[action])
## next_observation = numpy.copy(game.board)
num_moves_without_fruit += 1
if game.game_over:
reward = -10
elif game.just_ate_fruit:
reward = 1
elif num_moves_without_fruit > max_moves_without_fruit:
reward = -1
num_moves_without_fruit = 0
else:
reward = 0
memory.add_to_memory(observation, action, reward)
if game.game_over:
#### total_reward = sum(memory.rewards)
total_observation = numpy.stack(memory.observations, 0)
total_action = numpy.array(memory.actions)
total_rewards = discount_rewards(memory.rewards, gamma)
train_step(snake_model, optimizer, total_observation,
total_action, total_rewards)
memory.clear()
break
