Function ffn_gateup_fused_parts 

pub fn ffn_gateup_fused_parts(hidden: usize, inter: usize) -> (usize, usize)

ffn_gateup_fused_bytes split into (weight bytes, scale-grid bytes) — the loader adopts the two buffers separately, so it needs the terms rather than the sum, and taking them from here is what keeps the prediction and the tally one arithmetic.