我能够解决这个问题.

df['rolling_max'] = df.groupby(['Symbol',df.groupby('Symbol')['condition'].cumsum()])['number'].cummax()