dodox/fancy_gym
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

bug fixes

Browse Source
This commit is contained in:
Maximilian Huettenrauch
2021-04-23 12:47:55 +02:00
parent ba0b612868
commit c2db2f8064
6 changed files with 45 additions and 124 deletions
Download Patch File Download Diff File Expand all files Collapse all files
alr_envs/utils/make_env_helpers.py
-108
View File
@@ -27,111 +27,3 @@ def make_detpmp_env(**kwargs):
name = kwargs.pop("name")
_env = gym.make(name)
return DetPMPWrapper(_env, **kwargs)
# def _worker(index, env_fn, pipe, parent_pipe, shared_memory, error_queue):
# assert shared_memory is None
# env = env_fn()
# parent_pipe.close()
# try:
# while True:
# command, data = pipe.recv()
# if command == 'reset':
# observation = env.reset()
# pipe.send((observation, True))
# elif command == 'configure':
# env.configure(data)
# pipe.send((None, True))
# elif command == 'step':
# observation, reward, done, info = env.step(data)
# if done:
# observation = env.reset()
# pipe.send(((observation, reward, done, info), True))
# elif command == 'seed':
# env.seed(data)
# pipe.send((None, True))
# elif command == 'close':
# pipe.send((None, True))
# break
# elif command == '_check_observation_space':
# pipe.send((data == env.observation_space, True))
# else:
# raise RuntimeError('Received unknown command `{0}`. Must '
# 'be one of {`reset`, `step`, `seed`, `close`, '
# '`_check_observation_space`}.'.format(command))
# except (KeyboardInterrupt, Exception):
# error_queue.put((index,) + sys.exc_info()[:2])
# pipe.send((None, False))
# finally:
# env.close()
#
#
# def _worker_shared_memory(index, env_fn, pipe, parent_pipe, shared_memory, error_queue):
# assert shared_memory is not None
# env = env_fn()
# observation_space = env.observation_space
# parent_pipe.close()
# try:
# while True:
# command, data = pipe.recv()
# if command == 'reset':
# observation = env.reset()
# write_to_shared_memory(index, observation, shared_memory,
# observation_space)
# pipe.send((None, True))
# elif command == 'configure':
# env.configure(data)
# pipe.send((None, True))
# elif command == 'step':
# observation, reward, done, info = env.step(data)
# if done:
# observation = env.reset()
# write_to_shared_memory(index, observation, shared_memory,
# observation_space)
# pipe.send(((None, reward, done, info), True))
# elif command == 'seed':
# env.seed(data)
# pipe.send((None, True))
# elif command == 'close':
# pipe.send((None, True))
# break
# elif command == '_check_observation_space':
# pipe.send((data == observation_space, True))
# else:
# raise RuntimeError('Received unknown command `{0}`. Must '
# 'be one of {`reset`, `step`, `seed`, `close`, '
# '`_check_observation_space`}.'.format(command))
# except (KeyboardInterrupt, Exception):
# error_queue.put((index,) + sys.exc_info()[:2])
# pipe.send((None, False))
# finally:
# env.close()
# def viapoint_dmp(**kwargs):
# _env = gym.make("alr_envs:ViaPointReacher-v0")
# # _env = ViaPointReacher(**kwargs)
# return DmpWrapper(_env, num_dof=5, num_basis=5, duration=2, alpha_phase=2.5, dt=_env.dt,
# start_pos=_env.start_pos, learn_goal=False, policy_type="velocity", weights_scale=50)
#
#
# def holereacher_dmp(**kwargs):
# _env = gym.make("alr_envs:HoleReacher-v0")
# # _env = HoleReacher(**kwargs)
# return DmpWrapper(_env, num_dof=5, num_basis=5, duration=2, dt=_env.dt, learn_goal=True, alpha_phase=2,
# start_pos=_env.start_pos, policy_type="velocity", weights_scale=50, goal_scale=0.1)
#
#
# def holereacher_fix_goal_dmp(**kwargs):
# _env = gym.make("alr_envs:HoleReacher-v0")
# # _env = HoleReacher(**kwargs)
# return DmpWrapper(_env, num_dof=5, num_basis=5, duration=2, dt=_env.dt, learn_goal=False, alpha_phase=2,
# start_pos=_env.start_pos, policy_type="velocity", weights_scale=50, goal_scale=1,
# final_pos=np.array([2.02669572, -1.25966385, -1.51618198, -0.80946476, 0.02012344]))
#
#
# def holereacher_detpmp(**kwargs):
# _env = gym.make("alr_envs:HoleReacher-v0")
# # _env = HoleReacher(**kwargs)
# return DetPMPWrapper(_env, num_dof=5, num_basis=5, width=0.005, policy_type="velocity", start_pos=_env.start_pos,
# duration=2, post_traj_time=0, dt=_env.dt, weights_scale=0.25, zero_start=True, zero_goal=False)
alr_envs/utils/mp_env_async_sampler.py
+6 -6
View File
@@ -23,7 +23,7 @@ def split_array(ary, size):
split = [k * size for k in range(1, repeat)]
sub_arys = np.split(ary, split)
if n_samples % repeat != 0:
if n_samples % size != 0:
tmp = np.zeros_like(sub_arys[0])
last = sub_arys[-1]
tmp[0: len(last)] = last
@@ -42,8 +42,8 @@ def _flatten_list(l):
class AlrMpEnvSampler:
"""
An asynchronous sampler for MPWrapper environments. A sampler object can be called with a set of parameters and
returns the corresponding final obs, rewards, dones and info dicts.
An asynchronous sampler for non contextual MPWrapper environments. A sampler object can be called with a set of
parameters and returns the corresponding final obs, rewards, dones and info dicts.
"""
def __init__(self, env_id, num_envs, seed=0):
self.num_envs = num_envs
@@ -68,10 +68,10 @@ class AlrMpEnvSampler:
if __name__ == "__main__":
env_name = "alr_envs:HoleReacherDMP-v0"
env_name = "alr_envs:ALRBallInACupSimpleDMP-v0"
n_cpu = 8
dim = 30
n_samples = 20
dim = 15
n_samples = 10
sampler = AlrMpEnvSampler(env_name, num_envs=n_cpu)
alr_envs/utils/wrapper/mp_wrapper.py
+2 -1
View File
@@ -66,7 +66,7 @@ class MPWrapper(gym.Wrapper, ABC):
if self.post_traj_steps > 0:
trajectory = np.vstack([trajectory, np.tile(trajectory[-1, :], [self.post_traj_steps, 1])])
velocity = np.vstack([velocity, np.zeros(shape=(self.post_traj_steps, self.dmp.num_dimensions))])
velocity = np.vstack([velocity, np.zeros(shape=(self.post_traj_steps, self.mp.num_dimensions))])
# self._trajectory = trajectory
# self._velocity = velocity
@@ -76,6 +76,7 @@ class MPWrapper(gym.Wrapper, ABC):
# TODO: @Max Why do we need this configure, states should be part of the model
# TODO: Ask Onur if the context distribution needs to be outside the environment
# TODO: For now create a new env with each context
# self.env.configure(context)
obs = self.env.reset()
info = {}