Note

Ray 2.40 uses RLlib’s new API stack by default. The Ray team has mostly completed transitioning algorithms, example scripts, and documentation to the new code base.

If you’re still using the old API stack, see New API stack migration guide for details on how to migrate.

EnvRunner API#

rllib.env.env_runner.EnvRunner#

Construction and setup#

`EnvRunner`	Base class for distributed RL-style data collection from an environment.
`EnvRunner.make_env`	Creates the RL environment for this EnvRunner and assigns it to `self.env`.
`EnvRunner.make_module`	Creates the RLModule for this EnvRunner and assigns it to `self.module`.
`EnvRunner.get_spaces`	Returns a dict mapping ModuleIDs to 2-tuples of obs- and action space.
`EnvRunner.assert_healthy`	Checks that self.__init__() has been completed properly.

Sampling#

`EnvRunner.sample`	Returns experiences (of any form) sampled from this EnvRunner.
`EnvRunner.get_metrics`	Returns metrics (in any form) of the thus far collected, completed episodes.

Cleanup#

EnvRunner.stop

Releases all resources used by this EnvRunner.

Single-agent and multi-agent EnvRunners#

By default, RLlib uses two built-in subclasses of EnvRunner, one for single-agent, one for multi-agent setups. It determines based on your config, which one to use.

Check your config.is_multi_agent property to find out, which of these setups you have configured and see the docs on setting up RLlib multi-agent for more details.