Method build_inp_pos

subprojects/llama.cpp/src/llama-graph.cpp:1447–1458

}

ggml_tensor * llm_graph_context::build_inp_pos() const {
    auto inp = std::make_unique<llm_graph_input_pos>(hparams.n_pos_per_embd());

    auto & cur = inp->pos;

    cur = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, (int64_t)n_tokens*hparams.n_pos_per_embd());
    ggml_set_input(cur);

    res->add_input(std::move(inp));

    return cur;
}

ggml_tensor * llm_graph_context::build_inp_attn_scale() const {
    auto inp = std::make_unique<llm_graph_input_attn_temp>(hparams.n_attn_temp_floor_scale, hparams.f_attn_temp_scale, hparams.f_attn_temp_offset);