ray.rllib.env.multi_agent_episode.MultiAgentEpisode.add_env_reset#

MultiAgentEpisode.add_env_reset(*, observations: Dict[Any, Any], infos: Dict[Any, Any] | None = None) → None[source]#

Stores initial observation.

Parameters:

  • observations – A dictionary mapping agent IDs to initial observations. Note that some agents may not have an initial observation.

  • infos – A dictionary mapping agent IDs to initial info dicts. Note that some agents may not have an initial info dict. If not None, the agent IDs in infos must be a subset of those in observations meaning it would not be allowed to have an agent with an info dict, but not with an observation.