ray.rllib.algorithms.algorithm_config.AlgorithmConfig.get_multi_rl_module_spec#

AlgorithmConfig.get_multi_rl_module_spec(*, env: Any | gymnasium.Env | None = None, spaces: Dict[str, Tuple[gymnasium.Space, gymnasium.Space]] | None = None, inference_only: bool = False, policy_dict: Dict[str, PolicySpec] | None = None, single_agent_rl_module_spec: RLModuleSpec | None = None) → MultiRLModuleSpec[source]#

Returns the MultiRLModuleSpec based on the given env/spaces.

Parameters:

  • env – An optional environment instance, from which to infer the different spaces for the individual RLModules. If not provided, tries to infer from spaces, otherwise from self.observation_space and self.action_space. Raises an error, if no information on spaces can be inferred.

  • spaces – Optional dict mapping ModuleIDs to 2-tuples of observation- and action space that should be used for the respective RLModule. These spaces are usually provided by an already instantiated remote EnvRunner (call EnvRunner.get_spaces()). If not provided, tries to infer from env, otherwise from self.observation_space and self.action_space. Raises an error, if no information on spaces can be inferred.

  • inference_only – If True, the returned module spec is used in an inference-only setting (sampling) and the RLModule can thus be built in its light version (if available). For example, the inference_only version of an RLModule might only contain the networks required for computing actions, but misses additional target- or critic networks. Also, if True, the returned spec does NOT contain those (sub) RLModuleSpecs that have their learner_only flag set to True.

Returns:

A new MultiRLModuleSpec instance that can be used to build a MultiRLModule.