在GroupBy.apply中使用自定义lambda函数和DataFrame.to_dict，最后使用Series.to_dict：

d = (df.groupby('username')[['course','marks']]
       .apply(lambda x: x.to_dict('records'))
       .to_dict())

或者：

d = {name: g[['course','marks']].to_dict('records') for name, g in df.groupby('username')}

print (d)
{'alex': [{'course': 'physics', 'marks': 20}, 
          {'course': 'math', 'marks': 30},
          {'course': 'chemistry', 'marks': 40}], 
 'bob': [{'course': 'chemistry', 'marks': 25}]}

您可以使用下面的来获得所需的输出

import pandas as pd

# create the DataFrame
df = pd.DataFrame({
    'username': ['alex', 'bob', 'alex', 'alex'],
    'course': ['physics', 'chemistry', 'math', 'chemistry'],
    'marks': [20, 25, 30, 40]
})

# group the DataFrame by username
grouped = df.groupby('username')

# create a list of nested dictionaries
result_list = []
for name, group in grouped:
    # create a dictionary for the current username
    user_dict = {}
    for _, row in group.iterrows():
        # add the course and marks for the current row to the user_dict
        user_dict[row['course']] = {
            'course': row['course'],
            'marks': row['marks']
        }
    # add the user_dict to the result list
    result_list.append({name: list(user_dict.values())})

print(result_list)

输出：

[
    {
        'alex': [
            {'course': 'physics', 'marks': 20},
            {'course': 'math', 'marks': 30},
            {'course': 'chemistry', 'marks': 40}
        ]
    },
    {
        'bob': [
            {'course': 'chemistry', 'marks': 25}
        ]
    }
]