Method n_gqa

llama.cpp:1294–1296

    }

    uint32_t n_gqa() const {
        return n_head/n_head_kv;
    }

    uint32_t n_embd_head() const {
        return n_embd/n_head;