Enum ExpertRefreshPolicy 

pub enum ExpertRefreshPolicy {
    EveryForward,
    EveryDecodeSteps(usize),
    EveryDenoiseSteps(usize),
}

When to re-run hit counting and expert placement.

Variants

EveryForward

Refresh on every forward (τ = 1; Mixtral-Offload-style).

EveryDecodeSteps(usize)

Autoregressive decode: refresh every N generated tokens / steps.

EveryDenoiseSteps(usize)

Diffusion block decode: refresh every N denoise steps within a block (jump_steps in the TIDE reference repo).

Trait Implementations

impl Clone for ExpertRefreshPolicy

fn clone(&self) -> ExpertRefreshPolicy

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ExpertRefreshPolicy

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

Formats the value using the given formatter.

impl Default for ExpertRefreshPolicy

fn default() -> Self

Returns the “default value” for a type.

impl PartialEq for ExpertRefreshPolicy

fn eq(&self, other: &ExpertRefreshPolicy) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl Copy for ExpertRefreshPolicy

impl Eq for ExpertRefreshPolicy

impl StructuralPartialEq for ExpertRefreshPolicy