pub fn glm5_next_mtp_layer_index(config: &ModelConfig) -> usizeExpand description
num_hidden_layers counts text layers only; the MTP layer sits at index
num_hidden_layers (45) and is NOT included in that count.
🪤 GLM-5.3 does not use DeepSeek’s mtp.0.* naming — the MTP weights
live under model.language_model.layers.45.*. A grep mtp over this
checkpoint returns zero hits. Verified over all 120 shard headers.