cannot import name '_check_fit_params' from 'sklearn.utils.validation'

[gilbertblanson]

• Operating System: Windows 11
• Python version: 3.10
• summit version used: 0.8.9

Description

I was trying to run the Summit tutorial (https://gosummit.readthedocs.io/en/latest/tutorials/intro.html) locally on my computer, but ran into this issue when trying to import summit. I thought it might be an issue with my installation, so I tried it on the Colab, but it gave the same error.

What I Did

from summit import *
import matplotlib.pyplot as plt

ImportError                               Traceback (most recent call last)
Cell In[1], line 1
----> 1 from summit import *
      2 import matplotlib.pyplot as plt

File [~\AppData\Local\anaconda3\envs\summit_env\lib\site-packages\summit\__init__.py:19](http://localhost:8888/~/AppData/Local/anaconda3/envs/summit_env/lib/site-packages/summit/__init__.py#line=18)
     17 from summit.domain import *
     18 from summit.experiment import *
---> 19 from summit.run import *
     20 from summit.strategies import *
     21 from summit.benchmarks import *

File [~\AppData\Local\anaconda3\envs\summit_env\lib\site-packages\summit\run.py:3](http://localhost:8888/~/AppData/Local/anaconda3/envs/summit_env/lib/site-packages/summit/run.py#line=2)
      1 from summit.strategies import Strategy, strategy_from_dict
      2 from summit.experiment import Experiment
----> 3 from summit.benchmarks import *
      4 from summit.utils.multiobjective import pareto_efficient, hypervolume
      5 from summit import get_summit_config_path

File [~\AppData\Local\anaconda3\envs\summit_env\lib\site-packages\summit\benchmarks\__init__.py:3](http://localhost:8888/~/AppData/Local/anaconda3/envs/summit_env/lib/site-packages/summit/benchmarks/__init__.py#line=2)
      1 from .snar import SnarBenchmark
      2 from .test_functions import Himmelblau, Hartmann3D, ThreeHumpCamel, DTLZ2, VLMOP2
----> 3 from .experimental_emulator import *
      4 from .MIT import *

File [~\AppData\Local\anaconda3\envs\summit_env\lib\site-packages\summit\benchmarks\experimental_emulator.py:36](http://localhost:8888/~/AppData/Local/anaconda3/envs/summit_env/lib/site-packages/summit/benchmarks/experimental_emulator.py#line=35)
     30 from sklearn.model_selection._validation import (
     31     _fit_and_score,
     32     _score,
     33     _aggregate_score_dicts,
     34 )
     35 from sklearn.metrics import r2_score
---> 36 from sklearn.utils.validation import (
     37     _deprecate_positional_args,
     38     indexable,
     39     check_is_fitted,
     40     _check_fit_params,
     41 )
     42 from sklearn.utils import check_array, _safe_indexing
     43 from sklearn.utils.fixes import delayed

ImportError: cannot import name '_check_fit_params' from 'sklearn.utils.validation' (~\AppData\Local\anaconda3\envs\summit_env\lib\site-packages\sklearn\utils\validation.py)

[gilbertblanson]

Fixed it by downgrading scikit-learn from 1.6 to 1.3.2.

[Eleftheria14]

This is fixed in Summit Remastered v0.9.0 🎉

The issue was that scikit-learn 1.4+ removed _check_fit_params from the public API, causing an ImportError.

How we fixed it:
We added try/except blocks with fallback implementations for sklearn 1.4+ compatibility in summit/benchmarks/experimental_emulator.py (lines 37-52):

try:
    from sklearn.utils.validation import _check_fit_params
except ImportError:
    # _check_fit_params was removed in scikit-learn 1.4+
    # We'll implement our own version
    def _check_fit_params(X, fit_params, indices=None):
        # ... fallback implementation

Testing:

• Now compatible with scikit-learn 1.7+
• All 668 tests passing
• No import errors

To use the fixed version:

git clone https://github.com/Eleftheria14/summit.git
cd summit
uv sync --all-extras

See the main README for full installation instructions. Summit Remastered is a community-maintained fork that modernizes the original Summit with updated dependencies while maintaining 100% API compatibility.

[gilbertblanson]

@Eleftheria14 That's great to hear! However, I can't find your repo. Have you set it to private?