def test(times = 10):
model = load_model('my_model.h5')
s = []
for i in range(times):
print(i)
pygame.event.pump()
failed = False
lead_x = 70
lead_y = 70
snake = NeuralNetwork_Snake(showScreen, screenx, screeny, snakeImg, lead_x, lead_y)
apple = Apple(showScreen, screenx, screeny, b_size, appleImg, snake.snake_list)
while not failed:
apple_x, apple_y = apple.apple_pos()
snake.update_snake_list(apple_x, apple_y)
if snake.is_alive() is False:
failed = True
s.append(snake.snake_length)
showScreen.fill(white)
if snake.eaten is True:
apple.update_apple_pos(snake.snake_list)
apple_x, apple_y = apple.apple_pos()
allPredictions = {}
for act in moves:
state = snake.state([apple_x, apple_y], act)
allPredictions[act] = model.predict(np.array(state).reshape(-1, 5))[0][0]
act = max(allPredictions, key=allPredictions.get)
snake.set_direction(act)
apple.display()
snake.eaten = False
snake.display()
snake.display_score()
pygame.display.update()
time.tick(FPS)
print("Avg is: {}".format(sum(s)/len(s)))
def trainSnake(times = 40000):
train_data = []
for i in range(times):
print(i)
pygame.event.pump()
failed = False
lead_x = 70
lead_y = 70
snake = NeuralNetwork_Snake(showScreen, screenx, screeny, snakeImg, lead_x, lead_y)
apple = Apple(showScreen, screenx, screeny, b_size, appleImg, snake.snake_list)
apple_x, apple_y = apple.apple_pos()
act = "up"
state = snake.state([apple_x, apple_y], act)
former_distance = snake.distance([apple_x, apple_y])
while not failed:
apple_x, apple_y = apple.apple_pos()
snake.update_snake_list(apple_x, apple_y)
distance = snake.distance([apple_x, apple_y])
score = 0
if snake.is_alive() is False:
failed = True
score = -1
showScreen.fill(white)
if snake.eaten is True: apple.update_apple_pos(snake.snake_list)
if snake.eaten is True or distance < former_distance: score = 1
train_data.append([np.array(state), score])
act = random.choice(moves)
apple_x, apple_y = apple.apple_pos()
former_distance = snake.distance([apple_x, apple_y])
state = snake.state([apple_x, apple_y], act)
snake.set_direction(act)
apple.display()
snake.eaten = False
snake.display()
snake.display_score()
pygame.display.update()
time.tick(FPS)
print(len(train_data))
f = open("train_data.txt", "wb")
pickle.dump(train_data, f)
f.close()