Skip to main content

compute_decode_token_host

Function compute_decode_token_host 

Source
pub fn compute_decode_token_host<M: DecodeBackbone + ?Sized>(
    model: &M,
    cached_len: usize,
    tokens: &[u32],
    max_seq_len: usize,
    rope_inv_freq: Option<&[f64]>,
) -> DecodeTokenHost
Expand description

Compute one decode step’s host-side data — the RoPE tables and the stacked mask. Pure host math, no graph and no upload, shared by the build path (which bakes or uploads it) and every family’s rebind hook (which uploads it).