def obtain_exact_trajectories(self,
n_traj_per_worker,
agent_update,
env_update=None):
self._update_workers(agent_update, env_update)
batches = []
for worker in self._workers:
for _ in range(n_traj_per_worker):
batch = worker.rollout()
batches.append(batch)
return SkillTrajectoryBatch.concatenate(*batches)
def obtain_samples(self, itr, num_samples, agent_update, env_update=None,
skill=None):
self._update_workers(agent_update, env_update)
batches = []
completed_samples = 0
while True:
for worker in self._workers:
if not isinstance(worker, SkillWorker) and not isinstance(worker, KantWorker):
raise ValueError('Worker used by Local Skill Sampler class'
' must be a Skill/Kant Worker object, but got '
'{}'.format(type(worker)))
batch = worker.rollout(skill)
completed_samples += len(batch.actions)
batches.append(batch)
if completed_samples >= num_samples:
return SkillTrajectoryBatch.concatenate(*batches)
def _obtain_evaluation_samples(self, env, num_trajs=100):
paths = []
max_path_length = self.max_eval_path_length
if max_path_length is None:
max_path_length = self.max_path_length
# Use a finite length rollout for evaluation.
if max_path_length is None or np.isinf(max_path_length):
max_path_length = 1000
for _ in range(num_trajs):
path = self._rollout(env,
self.policy,
max_path_length=max_path_length,
deterministic=True)
paths.append(path)
return SkillTrajectoryBatch.from_trajectory_list(
self.env_spec, self.skills_num, paths)
def collect_rollout(self):
states = self._states
self._states = []
last_states = self._last_states
self._last_states = []
skills = self._skills
self._skills = []
actions = self._actions
self._actions = []
rewards = self._rewards
self._rewards = []
terminals = self._terminals
self._terminals = []
env_infos = self._env_infos
self._env_infos = defaultdict(list)
agent_infos = self._agent_infos
self._agent_infos = defaultdict(list)
for k, v in agent_infos.items():
if k == "dist":
continue
agent_infos[k] = np.asarray(v)
for k, v in env_infos.items():
env_infos[k] = np.asarray(v)
lengths = self._lengths
self._lengths = []
# print(np.asarray(skills))
return SkillTrajectoryBatch(
env_spec=self.env.spec,
num_skills=self._num_skills,
skills=np.asarray(skills).reshape((np.asarray(skills).shape[0], )),
states=np.asarray(states),
last_states=np.asarray(last_states),
actions=np.asarray(actions),
env_rewards=np.asarray(rewards), # env_rewards
terminals=np.asarray(terminals),
env_infos=dict(env_infos),
agent_infos=dict(agent_infos),
lengths=np.asarray(lengths, dtype='i'))