def run():
with open("results/winner-pickle-"+fn_results, 'rb') as f:
c = pickle.load(f)
print('loaded genome:')
print(c)
local_dir = os.path.dirname(__file__)
config_path = os.path.join(local_dir, 'config-recurrent-small')
config = neat.Config(neat.DefaultGenome, neat.DefaultReproduction,
neat.DefaultSpeciesSet, neat.DefaultStagnation,
config_path)
net = neat.nn.recurrent.RecurrentNetwork.create(c, config)
register()
env = gym.make("PuyoPuyoEndlessSmall-v2")
done = False
ob = env.reset()
ticks = 0
total_reward = 0
while True:
env.render()
time.sleep(0.5)
pieces_sum, field_sum = multiplyMatrices(ob[0], ob[1])
next_piece = pieces_sum[0]
inp_piece = np.ndarray.flatten(next_piece)
inp_field = np.ndarray.flatten(field_sum)
inputs = np.hstack([inp_piece, inp_field])
nn_output = net.activate(inputs)
action = np.argmax(nn_output)
ob, rew, done, info = env.step(action)
total_reward += rew
ticks += 1
if done:
break
print("Game played for ", ticks, " turns.")
print("Total score: ", total_reward+ticks)
if DRAW_NETS:
visualize.draw_net(config, c, view=True,
filename="results/winner-"+fn_results+".net")
visualize.draw_net(config, c, view=True,
filename="results/winner-"+fn_results+"-enabled.net",
show_disabled=False)
visualize.draw_net(config, c, view=True,
filename="results/winner-"+fn_results+"-pruned.net",
show_disabled=False, prune_unused=True)
# note: we only calculate the loss on the actions we've actually taken
actions = tf.cast(actions, tf.int32)
policy_loss = weighted_sparse_ce(actions,
logits,
sample_weight=advantages)
# entropy loss can be calculated via CE over itself
entropy_loss = kls.categorical_crossentropy(logits,
logits,
from_logits=True)
# here signs are flipped because optimizer minimizes
return policy_loss - self.params['entropy'] * entropy_loss
if __name__ == '__main__':
logging.getLogger().setLevel(logging.INFO)
summary_writer = tf.summary.create_file_writer(log_path)
'''
gpus = tf.config.experimental.list_physical_devices('GPU')
if gpus:
try:
tf.config.experimental.set_virtual_device_configuration(gpus[0], [tf.config.experimental.VirtualDeviceConfiguration(memory_limit=5120)])
except RuntimeError as e:
print(e)
'''
register()
env = gym.make('PuyoPuyoEndlessTsu-v2')
model = Model(num_actions=env.action_space.n)
agent = A2CAgent(model)
with summary_writer.as_default():
rewards_history = agent.train(env)
