Skip to main content

matmul

Function matmul 

Source
pub fn matmul(
    output: &mut [f32],
    input: &[f32],
    weight: &[f32],
    m: usize,
    k: usize,
    n: usize,
)
Expand description

General matrix multiply: output[m,n] = input[m,k] * weight^T[k,n]

For m=1 (single token), delegates to the optimized vector version.