evorl.envs.env

Module Contents

Classes

Env

Unified EvoRL Env API.

EnvAdapter

Base class for an environment adapter.

EnvState

State of the environment.

Data

EnvResetFn

EnvStepFn

API

class evorl.envs.env.Env[source]

Bases: abc.ABC

Unified EvoRL Env API.

abstract property action_space: evorl.envs.space.Space

Get the action space of the environment.

abstract property obs_space: evorl.envs.space.Space

Get the observation space of the environment.

abstract reset(key: chex.PRNGKey) → evorl.envs.env.EnvState[source]

Reset the environment to initial state.

abstract step(state: evorl.envs.env.EnvState, action: evorl.types.Action) → evorl.envs.env.EnvState[source]

Take a step in the environment.

class evorl.envs.env.EnvAdapter(env: Any)[source]

Bases: evorl.envs.env.Env

Base class for an environment adapter.

Convert envs from other packages to EvoRL’s Env API.

property unwrapped: Any
evorl.envs.env.EnvResetFn

None

class evorl.envs.env.EnvState[source]

Bases: evorl.types.PyTreeData

State of the environment.

Include all the data needed to represent the state of the environment.

Variables:

  • env_state – The internal state of the environment.

  • obs – The observation of the environment.

  • reward – The reward of the environment.

  • done – Whether the environment is done.

  • info – Extra info from the environment.

  • _internal – Extra internal data for the environment.

done: evorl.types.Done

None

env_state: evorl.types.EnvInternalState

None

info: evorl.types.PyTreeDict

‘pytree_field(…)’

obs: evorl.types.Observation

None

reward: evorl.types.Reward

None

evorl.envs.env.EnvStepFn

None