ray.rllib.core.rl_module.rl_module.RLModuleSpec#

class ray.rllib.core.rl_module.rl_module.RLModuleSpec(module_class: Type[RLModule] | None = None, observation_space: gymnasium.Space | None = None, action_space: gymnasium.Space | None = None, inference_only: bool = False, learner_only: bool = False, model_config: Dict[str, Any] | DefaultModelConfig | None = None, catalog_class: Type[Catalog] | None = None, load_state_path: str | None = None, model_config_dict: dict | int | None = None)[source]#

Utility spec class to make constructing RLModules (in single-agent case) easier.

Parameters:

module_class – The RLModule class to use.
observation_space – The observation space of the RLModule. This may differ from the observation space of the environment. For example, a discrete observation space of an environment, would usually correspond to a one-hot encoded observation space of the RLModule because of preprocessing.
action_space – The action space of the RLModule.
inference_only – Whether the RLModule should be configured in its inference-only state, in which those components not needed for action computing (for example a value function or a target network) might be missing. Note that inference_only=True AND learner_only=True is not allowed.
learner_only – Whether this RLModule should only be built on Learner workers, but NOT on EnvRunners. Useful for RLModules inside a MultiRLModule that are only used for training, for example a shared value function in a multi-agent setup or a world model in a curiosity-learning setup. Note that inference_only=True AND learner_only=True is not allowed.
model_config – The model config dict or default RLlib dataclass to use.
catalog_class – The Catalog class to use.
load_state_path – The path to the module state to load from. NOTE: This must be an absolute path.

PublicAPI (beta): This API is in beta and may change before becoming stable.

Methods

`as_multi_rl_module_spec`	Returns a MultiRLModuleSpec (`self` under DEFAULT_MODULE_ID key).
`build`	Builds the RLModule from this spec.
`from_dict`	Returns a single agent RLModule spec from a serialized representation.
`to_dict`	Returns a serialized representation of the spec.
`update`	Updates this spec with the given other spec.

Attributes

`action_space`
`catalog_class`
`inference_only`
`learner_only`
`load_state_path`
`model_config`
`model_config_dict`
`module_class`
`observation_space`