how to speed up saving the result to dataframe

[ZhangAllen98]

I am using the moirai to do zeroshot forecast，and I found that the speed of saving the result to dataframe is much slow.
Below is an example. First we get the forecasts, which a generator. And I using the following way to save the result as a dataframe, and when we use the moirai-large version, the time of saving result in my case would be about 77s, and when we use the moirai-small version ,the time of saving result in my case would be about 22s. I want to know is there a simple and fast method to save the result of the forecasts, and why the time of saving the result of different version would be different?

    t1 = time.time()
    pred_input = PandasDataset(dict(df_wide))
    forecasts = predictor.predict(pred_input)
    t2 = time.time()
    print(f'predict using {t2- t1} s ')
    # -------------------------------------------------

    output_list = []
    for i, temp_data in enumerate(forecasts):

        sample_array = temp_data.samples
        temp_df = pd.DataFrame({'value': np.median(sample_array, axis=0), 
                                'forecast_time': pd.date_range(temp_data.start_date.to_timestamp(),
                                                               periods=pred_len, freq='1h'),
                                'site': temp_data.item_id,
                                }
                               )
        output_list.append(temp_df)

    output_df = pd.concat(output_list, )
    t3 = time.time()
    print(f'save result using {t3 - t2} s ')