Struct SchedulerConfig 

pub struct SchedulerConfig {

    pub policy: SchedulingPolicy,
    pub max_waiting_requests: usize,
    pub max_running_requests: usize,
    pub enable_preemption: bool,
    pub enable_load_balancing: bool,
    pub fair_share_weights: HashMap<String, f32>,
    pub enable_sla_enforcement: bool,
    pub prompt_token_estimate: bool,
    pub prefill_first_until_active: Option<usize>,
    pub prefix_rendezvous_max_wait_ms: Option<NonZero<u64>>,
    pub prefill_step_chunk: Option<usize>,
    pub active_decode_prefill_chunk: Option<usize>,
    pub scheduler_none_prof: bool,
    pub sequence_fit_policy: SequenceFitPolicy,
}

Scheduler configuration

Fields

policy: SchedulingPolicy

Scheduling policy

max_waiting_requests: usize

Maximum waiting queue size

max_running_requests: usize

Maximum running requests

enable_preemption: bool

Enable request preemption

enable_load_balancing: bool

Enable load balancing

fair_share_weights: HashMap<String, f32>

Fair share weights per client

enable_sla_enforcement: bool

SLA enforcement enabled

prompt_token_estimate: bool

Use prompt-token metadata for initial continuous-batch admission estimates.

prefill_first_until_active: Option<usize>

Prefer new prefills over early decodes until this many requests are active.

prefix_rendezvous_max_wait_ms: Option<NonZero<u64>>

Opt in to one bounded wait for another request’s exact in-flight prefix. Expiry releases the optional dependency and resumes cold admission.

prefill_step_chunk: Option<usize>

Optional hard cap for per-request prefill chunks. None spends the live per-step token budget and lets capacity feedback narrow or regrow each request independently.

active_decode_prefill_chunk: Option<usize>

Cap prefill admission chunks only while decode requests are already active.

scheduler_none_prof: bool

Emit diagnostic scheduler None/SOME decisions.

sequence_fit_policy: SequenceFitPolicy

Non-reserving sequence fit gate used before prefill admission.

Implementations

impl SchedulerConfig

pub fn apply_runtime_config_snapshot( &mut self, snapshot: &RuntimeConfigSnapshot, ) -> Result<(), String>

Trait Implementations

impl Clone for SchedulerConfig

fn clone(&self) -> SchedulerConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for SchedulerConfig

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

Formats the value using the given formatter.

impl Default for SchedulerConfig

fn default() -> SchedulerConfig

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for SchedulerConfig

fn deserialize<__D>( __deserializer: __D, ) -> Result<SchedulerConfig, <__D as Deserializer<'de>>::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for SchedulerConfig

fn serialize<__S>( &self, __serializer: __S, ) -> Result<<__S as Serializer>::Ok, <__S as Serializer>::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.