Struct border_async_trainer::ActorManager

pub struct ActorManager<A, E, R, P>where
    A: Agent<E, R> + SyncModel,
    E: Env,
    P: StepProcessorBase<E>,
    R: ReplayBufferBase<PushedItem = P::Output>,{ /* private fields */ }

Manages Actors.

This struct handles the following requests:

• From the LearnerManager for updating the latest model info, stored in this struct.
• From the Actors for getting the latest model info.
• From the Actors for pushing sample batch to the LearnerManager.

Implementations

impl<A, E, R, P> ActorManager<A, E, R, P>where A: Agent<E, R> + SyncModel, E: Env, P: StepProcessorBase<E>, R: ReplayBufferBase<PushedItem = P::Output> + Send + 'static, A::Config: Send + 'static, E::Config: Send + 'static, P::Config: Send + 'static, R::PushedItem: Send + 'static, A::ModelInfo: Send + 'static,

pub fn build( config: &ActorManagerConfig, agent_configs: &Vec<A::Config>, env_config: &E::Config, step_proc_config: &P::Config, pushed_item_message_sender: Sender<PushedItemMessage<R::PushedItem>>, model_info_receiver: Receiver<(usize, A::ModelInfo)>, stop: Arc<Mutex<bool>> ) -> Self

Builds a ActorManager.

pub fn run(&mut self, guard_init_env: Arc<Mutex<bool>>)

Runs threads for Actors and a thread for sending samples into the replay buffer.

A thread will wait for the initial SyncModel::ModelInfo from AsyncTrainer, which blocks execution of Actor threads.

pub fn join(self) -> Vec<ActorStat>

Waits until all actors finish.

pub fn stop(&self)

Stops actor threads.

pub fn stop_and_join(self) -> Vec<ActorStat>

Stops and joins actors.