pub fn build_dense_mlp(
gpu: &dyn GpuBackend,
cfg: &Glm5NextMlpConfig,
tp_rank: usize,
full_inter: usize,
prefix: &str,
load: LoadFn<'_>,
) -> Result<Glm5NextDenseMlpWeights>Expand description
TP-slice and upload one BF16 SwiGLU MLP — a dense layer, or a routed layer’s shared expert.
full_inter is the tensor’s width in the checkpoint; the rank keeps full_inter / tp.