Method update_kv_cache

cactus/engine/engine_model.cpp:368–371

}

void Model::update_kv_cache(CactusGraph* gb, size_t seq_len) {
    kv_cache_.update_from_graph(gb, cache_k_output_nodes_, cache_v_output_nodes_,
                               seq_len, config_.num_layers);
}

void Model::remove_thinking_tokens(const std::vector<std::pair<size_t, size_t>>& ranges) {
    for (auto it = ranges.rbegin(); it != ranges.rend(); ++it)