Struct PackedQ2Weight 

pub struct PackedQ2Weight {
    pub weight: DevicePtr,
    pub n: u32,
    pub k: u32,
    pub group: u16,
}

Keep-packed ternary Q2_0 weight: a single contiguous buffer of raw PrismML block_q2_0 blocks ([fp16 d][group/4 bytes of 2-bit codes], value = (code-1)*d), row-major over [n, k]. The scale is INLINE (one fp16 per group of group elements) — there is no companion scale tensor, unlike NVFP4/FP8. Consumed by the native q2_0_gemv decode kernel, which reads the scale from each block. Built from a WeightDtype::PackedQ2_0 store tensor under ATLAS_GGUF_NATIVE_Q2=1; the buffer is owned by the WeightStore, so this struct only borrows the pointer (no free on drop).

Fields

weight: DevicePtr

Raw packed block_q2_0 bytes, n * (k/group) * (2 + group/4) long.

n: u32

Output rows (weight is [n, k]).

k: u32

Input columns (contraction dim).

group: u16

Group size (128 or 64) — elements per block / per inline scale.

Implementations

impl PackedQ2Weight

pub fn is_null(&self) -> bool

True if the backing buffer is NULL (unset placeholder).

Trait Implementations

impl Clone for PackedQ2Weight

fn clone(&self) -> PackedQ2Weight

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for PackedQ2Weight

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

Formats the value using the given formatter.

impl From<PackedQ2Weight> for QuantWeight

fn from(w: PackedQ2Weight) -> Self

Converts to this type from the input type.

impl Copy for PackedQ2Weight