ray.rllib.core.rl_module.rl_module.RLModule.output_specs_exploration
ray.rllib.core.rl_module.rl_module.RLModule.output_specs_exploration#
- RLModule.output_specs_exploration() Union[List[Union[str, Tuple[str, ...]]], NestedDict[Optional[Constraint]]] [source]#
Returns the output specs of the forward_exploration method.
Override this method to customize the output specs of the inference call. The default implementation requires the forward_exploration to reutn a dict that has
action_dist
key and its value is an instance ofDistribution
. This assumption must always hold.