pub struct RoutedExpertRequest<'a, T> {
pub layer: usize,
pub input: &'a T,
pub routes: &'a GroupSelection<T>,
pub pass: ExpertPass,
}Expand description
One architecture route batch submitted to a runtime expert provider.
Fields§
§layer: usizeGlobal decoder layer requesting experts.
input: &'a TFlattened token rows submitted to the selected experts.
routes: &'a GroupSelection<T>Backend-native selected expert IDs, scores, and weights.
pass: ExpertPassWhether this route batch belongs to prefill or decode.
Implementations§
Source§impl<T> RoutedExpertRequest<'_, T>
impl<T> RoutedExpertRequest<'_, T>
Sourcepub const fn parameter_bank_access(&self) -> ParameterBankAccess
pub const fn parameter_bank_access(&self) -> ParameterBankAccess
Projects architecture execution semantics into the storage workload class exposed to backend parameter-bank mechanisms.
Auto Trait Implementations§
impl<'a, T> Freeze for RoutedExpertRequest<'a, T>
impl<'a, T> RefUnwindSafe for RoutedExpertRequest<'a, T>
impl<'a, T> Send for RoutedExpertRequest<'a, T>
impl<'a, T> Sync for RoutedExpertRequest<'a, T>
impl<'a, T> Unpin for RoutedExpertRequest<'a, T>
impl<'a, T> UnsafeUnpin for RoutedExpertRequest<'a, T>
impl<'a, T> UnwindSafe for RoutedExpertRequest<'a, T>
Blanket Implementations§
Source§impl<T> BorrowMut<T> for Twhere
T: ?Sized,
impl<T> BorrowMut<T> for Twhere
T: ?Sized,
Source§fn borrow_mut(&mut self) -> &mut T
fn borrow_mut(&mut self) -> &mut T
Mutably borrows from an owned value. Read more