Struct AgentEnvAsyncLoop 

pub struct AgentEnvAsyncLoop<BT: Backend, RLC: RLComponentsTypes> { /* private fields */ }

Available on crate feature rl only.

An asynchronous agent/environement interface.

Implementations

impl<BT: Backend, RLC: RLComponentsTypes> AgentEnvAsyncLoop<BT, RLC>

pub fn new( env_init: RLC::EnvInit, agent: AsyncPolicy<RLC::Backend, <<RLC as RLComponentsTypes>::LearningAgent as PolicyLearner<<RLC as RLComponentsTypes>::Backend>>::InnerPolicy>, config: AsyncAgentEnvLoopConfig, transition_device: &Device<BT>, transition_sender: Option<Sender<TimeStep<BT, <RLC as RLComponentsTypes>::State, <RLC as RLComponentsTypes>::Action, <RLC as RLComponentsTypes>::ActionContext>>>, trajectory_sender: Option<Sender<Trajectory<BT, <RLC as RLComponentsTypes>::State, <RLC as RLComponentsTypes>::Action, <RLC as RLComponentsTypes>::ActionContext>>>, ) -> Self

Create a new asynchronous runner.

Arguments

• env_init - A function returning an environment instance.
• agent - An AsyncPolicy taking actions in the loop.
• config - An AsyncAgentEnvLoopConfig.
• transition_sender - Optional sender for transitions if you want to drive the requests from outside of the loop instance.
• trajectory_sender - Optional sender for trajectories if you want to drive the requests from outside of the loop instance.

Returns

An async Agent/Environement loop.

Trait Implementations

impl<BT, RLC> AgentEnvLoop<BT, RLC> for AgentEnvAsyncLoop<BT, RLC>

where BT: Backend, RLC: RLComponentsTypes,

fn run_steps( &mut self, num_steps: usize, processor: &mut RLEventProcessorType<RLC>, interrupter: &Interrupter, progress: &mut Progress, ) -> Vec<TimeStep<BT, <RLC as RLComponentsTypes>::State, <RLC as RLComponentsTypes>::Action, <RLC as RLComponentsTypes>::ActionContext>>

Run a certain number of timesteps.

fn run_episodes( &mut self, num_episodes: usize, processor: &mut RLEventProcessorType<RLC>, interrupter: &Interrupter, _progress: &mut Progress, ) -> Vec<Trajectory<BT, <RLC as RLComponentsTypes>::State, <RLC as RLComponentsTypes>::Action, <RLC as RLComponentsTypes>::ActionContext>>

Run a certain number of episodes.

fn update_policy(&mut self, update: RLC::PolicyState)

Update the runner’s agent.

fn policy(&self) -> RLC::PolicyState

Get the state of the runner’s agent.