def objectItems(self, spec=None):
return SequenceWrapper(self, self._histories.items(), 1)
def objectIds(self, spec=None):
return SequenceWrapper(self, self._histories.keys())
def objectValues(self, spec=None):
return SequenceWrapper(self, self._histories.values())
obs, reward, done, info = env.step(action)
if done:
print("end of the game", "reward=", reward)
break
def run_episode(env, model, **kwargs):
obs = env.reset(training=False, **kwargs)
done = False
step = 0
while not done:
step += 1
action, _state = model.predict(obs, deterministic=True)
obs, reward, done, info = env.step(action)
print('obs=', obs, 'reward=', reward, 'step=', step)
env = TwoDimEnv()
env = SequenceWrapper(env)
sequence_id = str(np.random.randint(10000))
log_dir = './log/'+sequence_id
tensor_dir = './tensor/'+sequence_id
env = Monitor(env, log_dir)
#env.test()
#check_env(env)
model = SAC('MlpPolicy', env, verbose=0, tensorboard_log=tensor_dir)
model.learn(EPISODE_LENGTH*TRAINING_LENGTH, n_eval_episodes=TRAINING_LENGTH)
run_episode(env, model)