Struct TernaryWeight 

pub struct TernaryWeight {
    pub data: Vec<PackedTritVec>,
    pub scales: Vec<f32>,
    pub shape: (usize, usize),
    pub group_size: usize,
}

Ternary weight representation with per-group scales.

Weights are quantized to {-1, 0, +1} using AbsMean quantization, with a scale factor stored per group.

Fields

data: Vec<PackedTritVec>

Packed ternary values (bitsliced storage). Shape: [out_features, in_features] flattened.

scales: Vec<f32>

Scale factors per group. For a weight matrix [out, in], scales has shape [out, in/group_size].

shape: (usize, usize)

Original shape [out_features, in_features].

group_size: usize

Group size used for quantization.

Implementations

impl TernaryWeight

pub const fn out_features(&self) -> usize

Get the output features dimension.

pub const fn in_features(&self) -> usize

Get the input features dimension.

pub fn sparsity(&self) -> f32

Calculate the sparsity (fraction of zeros).

pub fn memory_bytes(&self) -> usize

Memory size in bytes.

pub fn compression_ratio(&self) -> f32

Compression ratio vs FP32.

Trait Implementations

impl Clone for TernaryWeight

fn clone(&self) -> TernaryWeight

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for TernaryWeight

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for TernaryWeight

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for TernaryWeight

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.