Struct ExpertResidency 

pub struct ExpertResidency {
    pub packed_addr: [u64; 3],
    pub scale_addr: [u64; 3],
    pub scale2: [f32; 3],
    pub input_scale: [Option<f32>; 3],
}

The six sub-buffer device addresses (+ scalars) of one resident expert — exactly what the ptr-table patcher writes into the shadow tables.

Fields

packed_addr: [u64; 3]

gate/up/down B_packed device VA.

scale_addr: [u64; 3]

gate/up/down B_scale device VA.

scale2: [f32; 3]

gate/up/down per-tensor weight_scale_2.

input_scale: [Option<f32>; 3]

gate/up/down input_scale (None = weight-only W4A16).

Trait Implementations

impl Clone for ExpertResidency

fn clone(&self) -> ExpertResidency

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ExpertResidency

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

Formats the value using the given formatter.

impl PartialEq for ExpertResidency

fn eq(&self, other: &ExpertResidency) -> 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 ExpertResidency

impl StructuralPartialEq for ExpertResidency