pub fn forward_batch_graph(
model: &Arc<CmfModel>,
kv_id: u64,
layers: &[GraphLayer<'_>],
invf: &[f32],
h: &mut [f32],
nh: usize,
nkv: usize,
hd: usize,
rd: usize,
hidden: usize,
inter: usize,
positions: &[usize],
cap: usize,
gemma: bool,
eps: f32,
k: usize,
) -> boolExpand description
Batched prefill: k contiguous positions through the whole graph in one submit
(projections/FFN as GEMMs, attention/GDN looped over scratch). h is
[k·hidden] in/out; positions len k. wgpu only.