def play_one_time():
gamegrid = puzzle.GameGrid()
train_times = 0
NN_data = {}
model = obtain_model()
i = 0
while (1):
i += 1
# print("*"*25,"game_step: ", i, "*"*25)
NN_data_temp, event = MCTS.mcts_process(gamegrid.matrix, model)
if i == 1:
NN_data = NN_data_temp
else:
data_merge(NN_data, NN_data_temp)
gamegrid.action(event)
for l in gamegrid.matrix:
# print(l, event)
pass
if gamegrid.is_over:
score_tem = gamegrid.max_value
square_score = gamegrid.sum_square
break
return NN_data, score_tem, square_score
def explorer_game():
memory = []
prev_obs = []
choices = []
gamegrid = puzzle.GameGrid()
for _ in range(conf.goal_steps):
action, _ = explorer_move(
inv_puzzle.Game(gamegrid.score, gamegrid.first, gamegrid.matrix),
conf.options, 5) # Max depth == 5
done = gamegrid.move(action)
choices.append(action)
memory = [gamegrid.matrix, action]
gamegrid.update_idletasks()
gamegrid.update()
if done:
break
time.sleep(0.01)
gamegrid.destroy()
time.sleep(1)
return gamegrid.score, memory, choices
def visual_ia_game(model):
memory = []
prev_obs = []
choices = []
gamegrid = puzzle.GameGrid()
for _ in range(conf.goal_steps):
prev_obs = np.concatenate(gamegrid.matrix, axis=0)
prediction = sorted_prediction(
model.predict(prev_obs.reshape(-1, len(prev_obs), 1))[0])
for i in range(len(prediction)):
action = conf.options[prediction[i]]
done = gamegrid.move(action)
if not equals_grid(np.concatenate(gamegrid.matrix, axis=0),
prev_obs):
break
choices.append(action)
memory = [gamegrid.matrix, action]
gamegrid.update_idletasks()
gamegrid.update()
if done:
break
time.sleep(0.01)
gamegrid.destroy()
time.sleep(1)
return gamegrid.score, memory, choices
def manual_move():
gamegrid = puzzle.GameGrid()
training_data = []
game_memory = []
prev_observation = []
scores = []
accepted_scores = []
try:
training_data = np.load('saves/manual.npy').tolist()
i = training_data[-1]
training_data = training_data[:-1]
print('Continuing manual training')
except Exception as e:
print('First manual training, good luck!')
while not gamegrid.is_end_game:
prev_observation = np.concatenate(gamegrid.matrix, axis=0)
while not gamegrid.moved:
gamegrid.update_idletasks()
gamegrid.update()
gamegrid.moved = False
# Only if it is an active move, save it
if not equals_grid(np.concatenate(gamegrid.matrix, axis=0),
prev_observation):
if not gamegrid.can_undo:
# Redo this movement
game_memory.pop()
else:
action = gamegrid.last_move
game_memory.append([prev_observation, action])
if gamegrid.score >= conf.score_requirement:
accepted_scores.append(gamegrid.score)
for data in game_memory:
if data[1] == conf.options[0]:
output = [1, 0, 0, 0]
elif data[1] == conf.options[1]:
output = [0, 1, 0, 0]
elif data[1] == conf.options[2]:
output = [0, 0, 1, 0]
elif data[1] == conf.options[3]:
output = [0, 0, 0, 1]
training_data.append([data[0], output])
scores.append(gamegrid.score)
if training_data:
training_data.append(i + 1)
np.save('saves/manual.npy', np.array(training_data))
def playing(model):
gamegrid = puzzle.GameGrid()
i = 0
while (1):
NN_data_temp, event = MCTS.mcts_process(gamegrid.matrix, model)
gamegrid.action(event)
print("step: ", i)
i += 1
for l in gamegrid.matrix:
print(l, event)
if gamegrid.is_over:
score_tem = gamegrid.max_value
break
return score_tem
def playing(cpuct, times):
gamegrid = puzzle.GameGrid()
i = 0
while (1):
event = MCTS.mcts_process(gamegrid.matrix,
cpuct=cpuct,
update_times=times)
gamegrid.action(event)
# print("step: ", i)
i += 1
for l in gamegrid.matrix:
# print(l, event)
pass
if gamegrid.is_over:
score_tem = gamegrid.max_value
break
return score_tem
assert direction in all_directions
new_bits = {
dir_up: 0,
dir_down: 1,
dir_left: 2,
dir_right: 3,
}[direction]
self.random_state = (
(self.random_state << 2) | new_bits) & self.random_state_bitmask
def rand_mod(self, mod):
self.prng_step()
return self.random_state % mod
# takes a step in the pseudo random number generator
def prng_step(self):
# Does steps of of a fibonacci LFSR
# https://en.wikipedia.org/wiki/Linear-feedback_shift_register#Fibonacci_LFSRs
# This PRNG method was chosen due to it being simple to implement and not terrible.
lfsr = self.random_state
for i in range(self.random_state_bitwidth):
bit = 1 & ((lfsr >> 0) ^ (lfsr >> 2) ^ (lfsr >> 3) ^ (lfsr >> 5))
lfsr = (lfsr >> 1) | (bit << (self.random_state_bitwidth - 1))
assert lfsr == lfsr & self.random_state_bitmask
self.random_state = lfsr
if __name__ == '__main__':
import puzzle
puzzle.GameGrid()
def initializeGame(seed = None, showGUI = True):
return puzzle.GameGrid(is_ai_game=True, useSeed = seed, showGUI = showGUI)
def decimal(button4):
global y
import puzzle
puzzle.GameGrid()
tabla_puntaje(y)