Struct ActorThrottle 

pub struct ActorThrottle {
    pub steps_per_broadcast: usize,
    pub max_lead_steps: usize,
}

Per-actor staleness throttle derived from AsyncActorLearnerConfig::actor_throttle.

Policy version v proves the learner has consumed v * steps_per_broadcast transitions from this actor, so the actor pauses once steps_sent >= v * steps_per_broadcast + max_lead_steps and waits for the next broadcast. See AsyncActorLearnerConfig::max_lead_steps for why the budget is cumulative.

Fields

steps_per_broadcast: usize

Env steps per actor the learner consumes between successive policy broadcasts (broadcast_every * num_steps).

max_lead_steps: usize

Maximum steps the actor may run ahead of proven consumption. 0 disables the throttle entirely (unbounded lead — only safe when the caller guarantees consumption keeps up).

Implementations

impl ActorThrottle

pub fn disabled() -> Self

A disabled throttle (unbounded actor lead).

pub fn should_pause(&self, steps_sent: usize, version: u64) -> bool

Whether an actor that has sent steps_sent experiences and last loaded policy version should pause and wait for a broadcast.

Trait Implementations

impl Clone for ActorThrottle

fn clone(&self) -> ActorThrottle

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Copy for ActorThrottle

impl Debug for ActorThrottle

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.