attention

Function attention 

Source
pub fn attention(
    x: &[f32],
    seqlen: usize,
    start_pos: usize,
    w: &AttnWeights<'_>,
    c: &AttnCfg,
    fc: &[(f32, f32)],
    cache: &mut WindowCache,
) -> AttnRun
Expand description

2026-09-25: The ratio-0 attention forward. x: [seqlen][dim] bf16 values; fc from freqs_cis over the maximum sequence length. Writes the new kv rows into cache’s ring.