Method ensure_model_loaded

src/server/rest_handler.cpp:380–413

@brief Ensure the model is loaded @param model_tag the model tag

///@brief Ensure the model is loaded
///@param model_tag the model tag
bool RestHandler::ensure_model_loaded(const std::string& model_tag) {
    std::string ensure_tag = model_tag;
    if (current_model_tag != ensure_tag) {
        std::this_thread::sleep_for(std::chrono::milliseconds(500));
        if (auto_chat_engine != nullptr) {
            auto_chat_engine.reset();
        }
        std::pair<std::string, std::unique_ptr<AutoModel>> auto_model = get_auto_model(ensure_tag, this->supported_models, &this->npu_device_inst);
        auto_chat_engine = std::move(auto_model.second);
        ensure_tag = auto_model.first;
        if (!downloader.is_model_downloaded(ensure_tag)) {
            downloader.pull_model(ensure_tag);
        }
        auto [new_ensure_tag, model_info] = supported_models.get_model_info(ensure_tag);
        auto_chat_engine->configure_parameter("img_pre_resize", this->img_pre_resize);
        try {
            auto_chat_engine->load_model(supported_models.get_model_path(new_ensure_tag), model_info, ctx_length, preemption);
        }
        catch (const std::exception& e) {
            header_print("ERROR", "Failed to load model: " + std::string(e.what()));
            this->auto_chat_engine.reset();
            this->npu_device_inst.reset();
            this->npu_device_inst = xrt::device(0);
            this->current_model_tag = "model-faker";
            return false;
        }
        
        if (this->prefill_chunk_len == -1) {
            this->prefill_chunk_len = model_info["max_prefill_len"].get<int>();;
        }
        current_model_tag = ensure_tag;
    }
    return true;
}

///@brief Ensure the asr model is loaded
///@param model_tag the model tag