Method get_k

subprojects/llama.cpp/src/llama-kv-cache.cpp:1016–1034

}

ggml_tensor * llama_kv_cache::get_k(ggml_context * ctx, int32_t il, uint32_t n_kv, const slot_info & sinfo) const {
    const int32_t ikv = map_layer_ids.at(il);

    auto * k = layers[ikv].k;

    const uint64_t kv_size      = get_size();
    const uint64_t n_embd_k_gqa = k->ne[0];

    assert(n_embd_k_gqa == hparams.n_embd_k_gqa(il));

    const uint32_t ns = sinfo.s1 - sinfo.s0 + 1;

    return ggml_view_4d(ctx, k,
            hparams.n_embd_head_k, hparams.n_head_kv(il), n_kv, ns,
            ggml_row_size(k->type, hparams.n_embd_head_k),
            ggml_row_size(k->type, n_embd_k_gqa),
            ggml_row_size(k->type, n_embd_k_gqa*kv_size),
            ggml_row_size(k->type, n_embd_k_gqa*kv_size)*sinfo.s0);
}

ggml_tensor * llama_kv_cache::get_v(ggml_context * ctx, int32_t il, uint32_t n_kv, const slot_info & sinfo) const {
    const int32_t ikv = map_layer_ids.at(il);