tf_agents.bandits.policies.constraints.BaseConstraint

Abstract base class for representing constraints.

tf_agents.bandits.policies.constraints.BaseConstraint(
    time_step_spec: tf_agents.typing.types.TimeStep,
    action_spec: tf_agents.typing.types.BoundedTensorSpec,
    name: Optional[Text] = None
)

The constraint class provides feasibility computation functionality for computing the probability of actions being feasible.

time_step_spec A TimeStep spec of the expected time_steps.
action_spec A nest of BoundedTensorSpec representing the actions.
name Python str name of this constraint.

Methods

__call__

View source

@abc.abstractmethod
__call__(
    observation: tf_agents.typing.types.NestedTensor,
    actions: Optional[types.Tensor] = None
) -> tf_agents.typing.types.Tensor

Returns the probability of input actions being feasible.