Method insert

src/common/AutoModel/modeling_qwen3.cpp:60–100

}

bool Qwen3::insert(chat_meta_info_t& meta_info, lm_uniform_input_t& input, std::function<bool()> is_cancelled) {
    // preprocess
    this->profiler_list[TKOEN_ENCODE_TIME].start();
    std::string templated_text;
    if (input.messages.empty() && input.prompt.empty()) {
        header_print("WARNING", "No messages or prompt provided");
        return false;
    }
    if (!input.messages.empty()) { // already a formated messages, usually from REST API
        templated_text = this->apply_chat_template(input.messages, input.tools);
    }
    else if (!input.prompt.empty()) { // a pure text, usually from the cli
        nlohmann::ordered_json messages;

        messages.push_back({ {"role", "user"}, {"content", input.prompt} });
        templated_text = this->apply_chat_template(messages);
    }

    std::vector<int> tokens = this->tokenizer->encode(templated_text);

    this->profiler_list[TKOEN_ENCODE_TIME].stop(tokens.size());

    // hardware
    int restore_idx = -1;
    qwen3_npu *qwen3_engine = dynamic_cast<qwen3_npu*>(this->lm_engine.get());

    if (meta_info.restore_allowed) {
        restore_idx = qwen3_engine->restore();
        this->total_tokens = restore_idx;
        this->token_history = checkpoint_his; // restore the token history to be consistent with the restored KV cache, which is crucial for correct functioning of _shared_insert's prefix-matching logic
    }

    size_t n = tokens.size();
    tokens.resize(n - (this->enable_think ? 0 : 4));

    bool success = this->_shared_insert(meta_info, tokens, is_cancelled, nullptr);

    checkpoint_his = token_history;
    int checkpoint_idx = qwen3_engine->checkpoint();
    return success;
}

std::string Qwen3::generate(chat_meta_info_t& meta_info, int length_limit, std::ostream& os, std::function<bool()> is_cancelled) {
    std::vector<int> sampled_tokens;