def make_box(scalar_range_list, dtype, defaults):
bounds = [scalar_range2tuple(r, defaults) for r in scalar_range_list]
return Box(
low=np.array([b[0] for b in bounds], dtype=dtype),
high=np.array([b[1] for b in bounds], dtype=dtype),
dtype=dtype,
)
def __init__(self, index: int, spec: RewardSpace):
self.id = spec.name
self.index = index
self.range = scalar_range2tuple(spec.range)
self.success_threshold = (
spec.success_threshold if spec.has_success_threshold else None
)
self.deterministic = spec.deterministic
self.platform_dependent = spec.platform_dependent
def __init__(
self,
get_reward: Callable[[RewardRequest], Reward],
spaces: Sequence[RewardSpace],
):
self._get_reward = get_reward
self.session_id = -1
if not spaces:
raise ValueError("No reward spaces")
self.indices = {s.name: i for i, s in enumerate(spaces)}
self.ranges = {s.name: scalar_range2tuple(s.range) for s in spaces}
def make_seq(scalar_range, dtype, defaults):
return Sequence(
size_range=scalar_range2tuple(scalar_range, defaults),
dtype=dtype,
opaque_data_format=proto.opaque_data_format,
)