pub fn from_full_attention(
spec: &FullAttentionSpec,
names: &StageNames,
target: &SpatialTarget,
) -> Result<Vec<SpatialProgram>, SpatialReject>Expand description
C = full attention(Q, K, V) over t positions and h heads — t query rows of h heads, each reading its
own K/V head and attending over all t positions. Unmasked, as the vision tower’s block is: every row sees
every position.