pub enum ResidentMat {
Bf16(DevicePtr),
Q2K(DevicePtr),
Q3K(DevicePtr),
}Expand description
A resident [N, K] row-major projection as the GGUF path left it on the
device: expanded bf16, or the raw Q2_K / Q3_K blocks
(WeightDtype::Q2K / Q3K) for the K-quant GEMV (m <= 8, q8_1 rows) and
MMQ (tensor cores, D2S6 / D4 tiles) kernels.
Variants§
Implementations§
Source§impl ResidentMat
impl ResidentMat
Sourcepub fn at_rows(self, rows: usize, k: usize) -> ResidentMat
pub fn at_rows(self, rows: usize, k: usize) -> ResidentMat
The sub-matrix starting rows rows in (each row k weights long).
Trait Implementations§
Source§impl Clone for ResidentMat
impl Clone for ResidentMat
Source§fn clone(&self) -> ResidentMat
fn clone(&self) -> ResidentMat
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 ResidentMat
impl Debug for ResidentMat
impl Copy for ResidentMat
Auto Trait Implementations§
impl Freeze for ResidentMat
impl RefUnwindSafe for ResidentMat
impl Send for ResidentMat
impl Sync for ResidentMat
impl Unpin for ResidentMat
impl UnwindSafe for ResidentMat
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