pub struct Fp8ExpertWeight {
pub gate_proj: Fp8Weight,
pub up_proj: Fp8Weight,
pub down_proj: Fp8Weight,
}Expand description
FP8 expert weight: gate/up/down projections as FP8 block-scaled weights.
Used for native FP8 expert dispatch (no NVFP4 dequant overhead). Each projection stores [N, K] FP8 E4M3 weights + block scales.
Fields§
§gate_proj: Fp8Weight§up_proj: Fp8Weight§down_proj: Fp8WeightTrait Implementations§
Source§impl Clone for Fp8ExpertWeight
impl Clone for Fp8ExpertWeight
Source§fn clone(&self) -> Fp8ExpertWeight
fn clone(&self) -> Fp8ExpertWeight
Returns a duplicate of the value. Read more
1.0.0 · Source§fn clone_from(&mut self, source: &Self)
fn clone_from(&mut self, source: &Self)
Performs copy-assignment from
source. Read moreSource§impl Debug for Fp8ExpertWeight
impl Debug for Fp8ExpertWeight
Source§impl From<Fp8ExpertWeight> for QuantExpertWeight
impl From<Fp8ExpertWeight> for QuantExpertWeight
Source§fn from(w: Fp8ExpertWeight) -> Self
fn from(w: Fp8ExpertWeight) -> Self
Converts to this type from the input type.
impl Copy for Fp8ExpertWeight
Auto Trait Implementations§
impl Freeze for Fp8ExpertWeight
impl RefUnwindSafe for Fp8ExpertWeight
impl Send for Fp8ExpertWeight
impl Sync for Fp8ExpertWeight
impl Unpin for Fp8ExpertWeight
impl UnwindSafe for Fp8ExpertWeight
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