matthieu637 / ddrl

Deep Developmental Reinforcement Learning
MIT License
29 stars 3 forks source link

FetchReach env observation_spcae #5

Closed huangjiancong1 closed 5 years ago

huangjiancong1 commented 5 years ago
> /home/jim/ddrl/gym/run.py(174)<module>()
-> env = gym.make(env_name)
(Pdb) n
> /home/jim/ddrl/gym/run.py(175)<module>()
-> observation = env.reset()
(Pdb) p env
<TimeLimit<FetchReachEnv<FetchReach-v1>>>
(Pdb) n
> /home/jim/ddrl/gym/run.py(176)<module>()
-> nb_sensors = env.observation_space.shape[0]
(Pdb) p env.
env.__class__             env.__lt__                env._max_episode_steps
env.__delattr__           env.__module__            env._past_limit
env.__dict__              env.__ne__                env.action_space
env.__dir__               env.__new__               env.class_name
env.__doc__               env.__reduce__            env.close
env.__enter__             env.__reduce_ex__         env.compute_reward
env.__eq__                env.__repr__              env.env
env.__exit__              env.__setattr__           env.metadata
env.__format__            env.__sizeof__            env.observation_space
env.__ge__                env.__str__               env.render
env.__getattr__           env.__subclasshook__      env.reset
env.__getattribute__      env.__weakref__           env.reward_range
env.__gt__                env._elapsed_seconds      env.seed
env.__hash__              env._elapsed_steps        env.spec
env.__init__              env._episode_started_at   env.step
env.__le__                env._max_episode_seconds  env.unwrapped
(Pdb) p env.observation_space
Dict(achieved_goal:Box(3,), desired_goal:Box(3,), observation:Box(10,))
(Pdb) p env.observation_space.shape[0]
*** TypeError: 'NoneType' object is not subscriptable
(Pdb) p env.observation_space.shape

For each part in FetchReach's env.observation_space, which I need use as the observation_space in run.py?

matthieu637 commented 5 years ago

Fixed in 80c4f6669605e6ce5b876a3219fbb470e8f6b1e1.