You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Currently, using dynamic_action_space=True in conjunction with operators_as_actions=True for a domain does not correctly work for stochastic domains such as Triangle Tireworld in their original IPPC variants i.e. without the custom pddlgym mods.
if dynamic_action_space:
if self.domain.operators_as_actions and self._domain_is_strips:
self._action_space = LiteralActionSpace(
self.domain, self.action_predicates,
type_hierarchy=self.domain.type_hierarchy,
type_to_parent_types=self.domain.type_to_parent_types)
LiteralActionSpace uses FastDownward (FD) for initializing the action space and thus does not support stochastic domains.
The check should also check if sefl.domain.is_probabilistic is set to False.
Error
The error is usually parser errors such as undefined predicate (probabilistic which is understandable since FD tries to parse those keywords as predicates.
The text was updated successfully, but these errors were encountered:
Description
Currently, using
dynamic_action_space=True
in conjunction withoperators_as_actions=True
for a domain does not correctly work for stochastic domains such asTriangle Tireworld
in their original IPPC variants i.e. without the custom pddlgym mods.The issue is the following line of code (
pddlgym/pddlgym/core.py
Line 334 in ea2945d
LiteralActionSpace
uses FastDownward (FD) for initializing the action space and thus does not support stochastic domains.The check should also check if
sefl.domain.is_probabilistic
is set toFalse
.Error
The error is usually parser errors such as
undefined predicate (probabilistic
which is understandable since FD tries to parse those keywords as predicates.The text was updated successfully, but these errors were encountered: