Function generate_response

subprojects/llama.cpp/tools/mtmd/mtmd-cli.cpp:177–218

};

static int generate_response(mtmd_cli_context & ctx, int n_predict) {
    llama_tokens generated_tokens;
    for (int i = 0; i < n_predict; i++) {
        if (i > n_predict || !g_is_generating || g_is_interrupted) {
            LOG("\n");
            break;
        }

        llama_token token_id = common_sampler_sample(ctx.smpl, ctx.lctx, -1);
        generated_tokens.push_back(token_id);
        common_sampler_accept(ctx.smpl, token_id, true);

        if (llama_vocab_is_eog(ctx.vocab, token_id) || ctx.check_antiprompt(generated_tokens)) {
            LOG("\n");
            break; // end of generation
        }

        LOG("%s", common_token_to_piece(ctx.lctx, token_id).c_str());
        fflush(stdout);

        if (g_is_interrupted) {
            LOG("\n");
            break;
        }

        // eval the token
        common_batch_clear(ctx.batch);
        common_batch_add(ctx.batch, token_id, ctx.n_past++, {0}, true);
        if (llama_decode(ctx.lctx, ctx.batch)) {
            LOG_ERR("failed to decode token\n");
            return 1;
        }
    }

    std::string generated_text = common_detokenize(ctx.lctx, generated_tokens);
    common_chat_msg msg;
    msg.role    = "assistant";
    msg.content = generated_text;
    ctx.chat_history.push_back(std::move(msg));

    return 0;
}

static std::string chat_add_and_format(mtmd_cli_context & ctx, common_chat_msg & new_msg) {
    LOG_DBG("chat_add_and_format: new_msg.role='%s', new_msg.content='%s'\n",