spark_model::layers::glm5next_mlp::weights

Struct Glm5NextDenseMlpWeights

pub struct Glm5NextDenseMlpWeights {
    pub gate_proj: DevicePtr,
    pub up_proj: DevicePtr,
    pub down_proj: DevicePtr,
}

Expand description

A dense SwiGLU MLP in BF16 — layers 0..first_k_dense_replace, and the shared expert of every routed layer. Same shape, same kernels, different widths.

TP: gate_proj/up_proj are column-parallel (row-sliced, since each is stored [inter, hidden]), down_proj is row-parallel (column-sliced on [hidden, inter]). The output is therefore a partial sum whenever tp_world_size > 1.

Fields§

§gate_proj: DevicePtr

[local_inter, hidden] BF16.

§up_proj: DevicePtr

[local_inter, hidden] BF16.

§down_proj: DevicePtr

[hidden, local_inter] BF16 — row-parallel.

Glm5NextDenseMlpWeights

Struct Glm5NextDenseMlpWeights Copy item path

Fields§

Trait Implementations§

impl Clone for Glm5NextDenseMlpWeights

fn clone(&self) -> Glm5NextDenseMlpWeights

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

impl Debug for Glm5NextDenseMlpWeights

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

impl Copy for Glm5NextDenseMlpWeights

Auto Trait Implementations§

impl Freeze for Glm5NextDenseMlpWeights

impl RefUnwindSafe for Glm5NextDenseMlpWeights

impl Send for Glm5NextDenseMlpWeights

impl Sync for Glm5NextDenseMlpWeights

impl Unpin for Glm5NextDenseMlpWeights

impl UnwindSafe for Glm5NextDenseMlpWeights

Blanket Implementations§

impl<T> Any for Twhere T: 'static + ?Sized,

fn type_id(&self) -> TypeId

impl<T> Borrow<T> for Twhere T: ?Sized,

fn borrow(&self) -> &T

impl<T> BorrowMut<T> for Twhere T: ?Sized,

fn borrow_mut(&mut self) -> &mut T

impl<T> CloneToUninit for Twhere T: Clone,

unsafe fn clone_to_uninit(&self, dest: *mut u8)

impl<T> From<T> for T

fn from(t: T) -> T

impl<T> Instrument for T

fn instrument(self, span: Span) -> Instrumented<Self>

fn in_current_span(self) -> Instrumented<Self>

impl<T, U> Into<U> for Twhere U: From<T>,

fn into(self) -> U

impl<T> ToOwned for Twhere T: Clone,

type Owned = T

fn to_owned(&self) -> T

fn clone_into(&self, target: &mut T)

impl<T, U> TryFrom<U> for Twhere U: Into<T>,

type Error = Infallible

fn try_from(value: U) -> Result<T, <T as TryFrom<U>>::Error>

impl<T, U> TryInto<U> for Twhere U: TryFrom<T>,

type Error = <U as TryFrom<T>>::Error

fn try_into(self) -> Result<U, <U as TryFrom<T>>::Error>

impl<V, T> VZip<V> for Twhere V: MultiLane<T>,

fn vzip(self) -> V

impl<T> WithSubscriber for T

fn with_subscriber<S>(self, subscriber: S) -> WithDispatch<Self>where S: Into<Dispatch>,

fn with_current_subscriber(self) -> WithDispatch<Self>

Struct Glm5NextDenseMlpWeights

impl<T> Any for T
where T: 'static + ?Sized,

impl<T> Borrow<T> for T
where T: ?Sized,

impl<T> BorrowMut<T> for T
where T: ?Sized,

impl<T> CloneToUninit for T
where T: Clone,

impl<T, U> Into<U> for T
where U: From<T>,

impl<T> ToOwned for T
where T: Clone,

impl<T, U> TryFrom<U> for T
where U: Into<T>,

impl<T, U> TryInto<U> for T
where U: TryFrom<T>,

impl<V, T> VZip<V> for T
where V: MultiLane<T>,

fn with_subscriber<S>(self, subscriber: S) -> WithDispatch<Self>
where S: Into<Dispatch>,