Struct QuantizedMatmulIdParams 

pub struct QuantizedMatmulIdParams {
    pub m: u32,
    pub k: u32,
    pub n: u32,
    pub group_size: u32,
    pub bits: u32,
    pub n_expert_used: u32,
    pub num_experts: u32,
}

Parameters describing the expert-routed quantized matmul dimensions.

Fields

m: u32

Number of input rows (tokens).

k: u32

Inner dimension (shared between input and weight).

n: u32

Number of output columns per expert.

group_size: u32

Number of consecutive values sharing one scale/bias pair.

bits: u32

Quantization bit width (4, 6, or 8).

n_expert_used: u32

Number of experts each token is routed to (top-k).

num_experts: u32

Total number of experts in the weight tensor.

Trait Implementations

impl Clone for QuantizedMatmulIdParams

fn clone(&self) -> QuantizedMatmulIdParams

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for QuantizedMatmulIdParams

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

Formats the value using the given formatter.

impl Copy for QuantizedMatmulIdParams