Tricks to make saving images quicker?

pre_outcome_chart 関数から始め、10k-image ワークロードで pd.read_csv、fplt.plot、plt.savefig にかかる時間を測定します。現在のループと提案されている multiprocessing アプローチを比較し、生成されるチャートを変更せずに、保存にかかる合計時間が測定可能かつ再現可能な形で短縮されることを完了の定義とします。

I am using a friends script (so I don't know all the details in this script), but I wonder if there are simple tricks I can do here to make it save quicker.
I want to be able to save around 10k images.

I plan to incorporate multiprocessing to make it even quicker.

Thank you.

def pre_outcome_chart(month, save_dir, sample_df, test = True):
     if test:
          sample_df = sample_df.sample(100)
    
     for index, row in sample_df.iterrows():
            ticker = row['ticker']
            input_date = row['entry_date']
          
            #nput_date_string = input_date.strftime('%Y-%m-%d')
            startDate = pd.to_datetime(input_date).date() - relativedelta(months=month)
            endDate = pd.to_datetime(input_date).date() + relativedelta(days =1)
            startDateString = startDate.strftime('%Y%m%d')
            endDateString = endDate.strftime('%Y%m%d')
         
            filename = os.path.join(DATA_DIR, ticker + '.csv')
            df = pd.read_csv(filename, index_col=0, parse_dates=True)
            dt_range = pd.date_range(start=startDate, end=endDate)
            df = df[df.index.isin(dt_range)]
            # Drop any rows with missing data
            df.dropna(inplace=True)

            # Compute moving averages
            df['SMA50'] = df['Close'].rolling(window=50, min_periods=1).mean()
            df['SMA100'] = df['Close'].rolling(window=100, min_periods=1).mean()
            df['SMA200'] = df['Close'].rolling(window=200, min_periods=1).mean()

            # Define addplots
            ema10 = fplt.make_addplot(df['Close'].ewm(span=10, min_periods=1).mean(), color='#CBC3E3')
            ema20 = fplt.make_addplot(df['Close'].ewm(span=20, min_periods=1).mean(), color='#87CEEB')
            sma50 = fplt.make_addplot(df['SMA50'], color='red')
            sma100 = fplt.make_addplot(df['SMA100'], color='yellow')
            sma200 = fplt.make_addplot(df['SMA200'], color='white')

            # Define market colors and style
            mc = fplt.make_marketcolors(up='black',down='#f76757',
                                       edge={'up': '#13eda4', 'down': '#f76757'},
                                       wick={'up': '#13eda4', 'down': '#f76757'},
                                       volume={'up': '#13eda4', 'down': '#f76757'},
                                        )

            s = fplt.make_mpf_style(marketcolors=mc,facecolor='black',figcolor='black',
                                    gridcolor='gray',
                                    gridstyle='dotted',
                                    rc={'xtick.color':'white',
                                      'ytick.color':'white',
                                      'axes.labelcolor':'white',
                                      'text.color':'white',
                                      'axes.edgecolor': 'gray',
                                      'grid.alpha': 0.7,
                                      'grid.linewidth': 0.5,
                                    })



            # Plot the data
            fig, axlist = fplt.plot(
                df,
                type='candle',
                addplot=[ema10, ema20, sma50, sma100, sma200],
                style=s,
                figsize=(12,6),
                update_width_config={'candle_linewidth':1.0, 'candle_width':0.525, 'volume_width': 0.525},
                tight_layout=True,
                volume=True,
                ylabel='Price',
                xrotation=0,
                returnfig=True
            )

            # Add legend
            ax = axlist[0]
            legend_handles = [ax.lines[i] for i in range(len(ax.lines))]
            legend_labels = ['EMA10', 'EMA20', 'SMA50', 'SMA100', 'SMA200']
            ax.legend(legend_handles, legend_labels)

            # Get the index of the input_date candle
            input_date_index = df.index.get_loc(pd.to_datetime(input_date, format='%Y-%m-%d').floor('D'))

            # Get the coordinates of the input_date candle
            x_coord = input_date_index
            y_coord = df['High'].iloc[input_date_index]

            # Calculate the maximum value of the chart
            chart_max = max(df['High'].max(), df[['SMA50', 'SMA100', 'SMA200']].max().max())

            # Calculate the maximum range of the chart
            chart_range = df['High'].max() - df['Low'].min()

            # Calculate the desired arrow length as a fraction of the maximum range
            arrow_length_fraction = 0.1  # Adjust this value to control the arrow length
            arrow_length = arrow_length_fraction * chart_range

            # Calculate the desired gap size
            gap_size = arrow_length

            # Adjust the ylim to create the gap
            axlist[0].set_ylim(df['Low'].min(), 1.2 * df['High'].max() + gap_size)

            # Annotate the input_date above the input_date candle
            axlist[0].annotate('',
                               xy=(x_coord, y_coord),
                               xytext=(x_coord, y_coord+arrow_length),
                               arrowprops=dict(arrowstyle='->', color='yellow', linewidth=3),
                               color='yellow', ha='center', va='bottom')

            # Add the date tag at the top of the chart
            ax.annotate(ticker + '  '  + pd.to_datetime(input_date).date().strftime('%Y-%m-%d'),
                        xy=(x_coord, y_coord),
                        xytext=(x_coord, chart_max+arrow_length),
                        color='yellow', ha='center', va='bottom')
            # Save chart
           
            plt.savefig(
                f"{save_dir}/"
                + f"{ticker}_"
                + input_date
                + 'M_'
                 + '.png', dpi=300, bbox_inches='tight')
            plt.close(fig)`