robopal.wrappers.human_demo_wrapper module

class robopal.wrappers.human_demo_wrapper.Collection(num_samples: int = 0, model_file: str = None, states: List[numpy.ndarray] = <factory>, actions: List[numpy.ndarray] = <factory>, rewards: List[numpy.ndarray] = <factory>, dones: List[numpy.ndarray] = <factory>, obs: Dict[str, List[numpy.ndarray]] = <factory>, next_obs: Dict[str, List[numpy.ndarray]] = <factory>)

基类：object

actions: List[ndarray]

dones: List[ndarray]

model_file: str = None

next_obs: Dict[str, List[ndarray]]

num_samples: int = 0

obs: Dict[str, List[ndarray]]

rewards: List[ndarray]

states: List[ndarray]

class robopal.wrappers.human_demo_wrapper.HumanDemonstrationWrapper(env: ManipulateEnv, *, device: BaseDevice = None, collections_dir: str = '/home/docs/checkouts/readthedocs.org/user_builds/robopal/checkouts/latest/docs/source/../../robopal/collections/collections_1748360434_2707884', collect_freq=1, max_collect_horizon=100, is_drop_unsuccess_exp=True, is_drop_invalid_action=True, saved_action_type='velocity', is_render_actions=False)

基类：object

close()

get_action()

compute next actions from the device outputs

reset(seed=None, options=None)

save_collection()

Save the collected data in the hdf5 file.

step(action: ndarray | Dict[str, ndarray])

Input actions should be normalized, since the action will un-normalize before applying to the environment.