Function gdn_decode_f32_norm_snap 

pub fn gdn_decode_f32_norm_snap(
    gpu: &dyn GpuBackend,
    kernel: KernelHandle,
    h_state: DevicePtr,
    query: DevicePtr,
    key: DevicePtr,
    value: DevicePtr,
    gate: DevicePtr,
    beta: DevicePtr,
    z_gate: DevicePtr,
    norm_weight: DevicePtr,
    output: DevicePtr,
    h_inter: DevicePtr,
    batch_size: u32,
    num_k_heads: u32,
    num_v_heads: u32,
    k_dim: u32,
    v_dim: u32,
    eps: f32,
    stream: u64,
) -> Result<()>

super::gdn_decode_f32_norm + inline h-state snapshot.

h_inter receives the post-update (post-state-norm-clamp) H — the same bits left in h_state — or is skipped when NULL (the final verify position, whose snapshot index has no reader). Same grid/block and argument order as the parent, with h_inter appended.