Function attend_int8 

pub fn attend_int8(
    fp: FpUnit,
    q: &[f32],
    keys: &[(u16, &[i8])],
    values: &[(u16, &[i8])],
    scale: f32,
    words: Words,
    scaling: ActScaling,
) -> Vec<f32>

One query head’s attention over its cached int8 rows on the array: the query quantized as words and scaling name, each position’s int8_score times scale, the softmax, the weighted_int8 sum.