Python 大Pandas 每月重新抽样200万只和300万只

我有这个代码来计算回报:

import yfinance as yf
import numpy as np
import pandas as pd

df = yf.download('SPY', '2023-01-01')
df = df[['Close']]
df['d_returns'] = np.log(df.div(df.shift(1)))
df.dropna(inplace = True)

df_1M = pd.DataFrame()
df_2M = pd.DataFrame()
df_3M = pd.DataFrame()

df_1M['1M cummreturns'] = df.d_returns.cumsum().apply(np.exp)
df_2M['2M cummreturns']= df.d_returns.cumsum().apply(np.exp)
df_3M['3M cummreturns'] = df.d_returns.cumsum().apply(np.exp)

df1 = df_1M[['1M cummreturns']].resample('1M').max()
df2 = df_2M[['2M cummreturns']].resample('2M').max()
df3 = df_3M[['3M cummreturns']].resample('3M').max()

df1 = pd.concat([df1, df2, df3], axis=1)
df1

这提供了以下内容:

            1M cummreturns  2M cummreturns  3M cummreturns
Date            
2023-01-31  1.067381        1.067381        1.067381
2023-02-28  1.094428        NaN             NaN
2023-03-31  1.075022        1.094428        NaN
2023-04-30  1.092196        NaN             1.094428
2023-05-31  1.103356        1.103356        NaN
2023-06-30  1.164014        NaN             NaN
2023-07-31  1.202116        1.202116        1.202116
2023-08-31  1.198677        NaN             NaN
2023-09-30  1.184785        1.198677        NaN
2023-10-31  1.145738        NaN             1.198677
2023-11-30  1.198466        1.198466        NaN
2023-12-31  1.251746        NaN             NaN
2024-01-31  1.290032        1.290032        1.290032
2024-02-29  1.334174        NaN             NaN
2024-03-31  1.346699        1.346699        NaN
2024-04-30  NaN             NaN             1.346699

如何获取每行2M cummreturns列和3M cummreturns列中的有效值？

例如，2023-02-28行表示Feb-2023个月.列2M cummreturns和3M cummreturns需要在接下来的2个月和3个月的时间内分别从Feb-2023开始有最大回报，同样的方式1M cummreturns在接下来的1个月时间内给出最大回报.