sdpa_one

Function sdpa_one 

Source
pub fn sdpa_one(
    q: &[f32],
    k: &[f32],
    v: &[f32],
    t: usize,
    heads: usize,
    dq: usize,
    dv: usize,
) -> Vec<f32>
Expand description

Decode SDPA: one query [H, dq] against cached K/V of length t.