[fix] keep group_keys=False in Average Ensemble

[lingbai-kong]

Description

Explicitly passing the group_keys param in AverageEnsemble

• Before

results = results.groupby("datetime").apply(lambda df: (df - df.mean()) / df.std())

• After

results = results.groupby("datetime", group_keys=False).apply(lambda df: (df - df.mean()) / df.std())

Motivation and Context

The related issues is #1893. And this PR is similar to #1898

When we use the OnlineManager with AverageEnsemble, the The name datetime occurs multiple times raises. I believe this caused by the new feature in Pandas 2.0.0, where release notice says "group_keys now defaults to True."

How Has This Been Tested?

• [] Pass the test by running: pytest qlib/tests/test_all_pipeline.py under upper directory of qlib.
• If you are adding a new feature, test on your own test scripts.

Screenshots of Test Results (if appropriate):

1. Pipeline test:
2. Your own tests:

• before
  
• after
  

Types of changes

• Fix bugs
• Add new feature
• Update documentation

[SunsetWolf]

Hi, @lingbai-kong
I want to know how to test this pull request, can you guide me?

[lingbai-kong]

@SunsetWolf Sure, you can test with this script:

import pandas as pd
import numpy as np

dates = pd.date_range('2025-01-02', '2025-04-01', freq='B')
instruments = [f'SH{str(i).zfill(6)}' for i in range(600006, 600034)] + \
              [f'SZ{str(i).zfill(6)}' for i in range(301558, 301607)]
index = pd.MultiIndex.from_product(
    [dates, instruments],
    names=['datetime', 'instrument']
)
data = np.random.randn(len(index))
data = pd.Series(data, index=index)
print(data)

from qlib.model.ens.ensemble import AverageEnsemble
collect_dict = {}
cdd = collect_dict.setdefault('test', {})
cdd['model'] = data
AverageEnsemble()(collect_dict)

[SunsetWolf]

I'm happy to see your reply, I tested it with the script you provided, and this PR does solve the problem of multi index after pandas.groupby().
I have some more questions for you:
I saw this PR mentioned in issue 1893, what is the connection between them? Does this PR solve the problem mentioned in issue 1893?
Also, the main branch already solves the problem in the CI, you can merge the main branch to make the CI pass.

[lingbai-kong]

@SunsetWolf This PR doesn't directly fix #1893, PR (#1898) fixed it. I mentioned the issue just because they are same kind of problem: the newest Pandas makes group_keys default to True.

[SunsetWolf]

Hi, @lingbai-kong
It looks great now, thanks for your contribution to qlib.