[BUG] Too many values to unpack with `sklearn 1.8`

[loiseaujc]

Hej guys,

It seems like sklearn has changed something between 1.7 and 1.8 (eventhough I couldn't find it in the changelog) about how _preprocess_data behaves. The following code now raises a Too many values to unpack while it worked flawlessly with sklearn 1.7.2

Reproducing code example:

import numpy as np
import scipy as sp
import pysindy as ps
import sklearn as sk

from scipy.integrate import solve_ivp

def lorenz(t, u, p):
    #> Unpack variables.
    x, y, z = u
    #> Unpack parameters.
    σ, ρ, β = p
    #> Equations.
    dx = σ*(y - x)
    dy = x*(ρ-z) - y
    dz = x*y - β*z
    return dx, dy, dz

#> Parameters for the simulation.
p = σ, ρ, β = 10.0, 28.0, 8.0/3.0
u0 = np.array([1.0, 1.0, 1.0])
t = np.linspace(0, 100, 10000) ; Δt = t[1] - t[0]
tspan = (t.min(), t.max())

#> Run the simulation.
solution = solve_ivp(
    lambda t, u : lorenz(t, u, p),
    tspan,
    u0,
    t_eval = t
)

#> Divise les données entre jeu de test et jeu d'entraînement.
n = len(solution.y[0]) # Nombre de points de données.
x_train = solution.y[:, :3*n//4].T
x_test = solution.y[:, 3*n//4:].T
t_test = t[3*n//4:]

#> Création de l'estimateur SINDy.
model = ps.SINDy(
    optimizer = ps.STLSQ(), # Choix de l'optimisateur.
    feature_library = ps.PolynomialLibrary(degree=5) # Choix de la base de fonctions.
)

#> Entraînement du modèle.
model.fit(
    x_train, # Jeu d'entraînement.
    t = Δt, # Pas de temps.,
    feature_names=["x", "y", "z"]
)

Error message:

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
Cell In[7], line 14
      8 model = ps.SINDy(
      9     optimizer = ps.STLSQ(), # Choix de l'optimisateur.
     10     feature_library = ps.PolynomialLibrary(degree=5) # Choix de la base de fonctions.
     11 )
     13 #> Entraînement du modèle.
---> 14 model.fit(
     15     x_train, # Jeu d'entraînement.
     16     t = Δt, # Pas de temps.,
     17     feature_names=["x", "y", "z"]
     18 )

File [~/miniconda3/envs/pysindy/lib/python3.14/site-packages/pysindy/pysindy.py:382](http://localhost:46683/home/loiseau/miniconda3/envs/pysindy/lib/python3.14/site-packages/pysindy/pysindy.py#line=381), in SINDy.fit(self, x, t, x_dot, u, feature_names)
    380 x_dot = concat_sample_axis(x_dot)
    381 self.model = Pipeline(steps)
--> 382 self.model.fit(x, x_dot)
    383 self._fit_shape()
    385 return self

File [~/miniconda3/envs/pysindy/lib/python3.14/site-packages/sklearn/base.py:1336](http://localhost:46683/home/loiseau/miniconda3/envs/pysindy/lib/python3.14/site-packages/sklearn/base.py#line=1335), in _fit_context.<locals>.decorator.<locals>.wrapper(estimator, *args, **kwargs)
   1329     estimator._validate_params()
   1331 with config_context(
   1332     skip_parameter_validation=(
   1333         prefer_skip_nested_validation or global_skip_validation
   1334     )
   1335 ):
-> 1336     return fit_method(estimator, *args, **kwargs)

File [~/miniconda3/envs/pysindy/lib/python3.14/site-packages/sklearn/pipeline.py:621](http://localhost:46683/home/loiseau/miniconda3/envs/pysindy/lib/python3.14/site-packages/sklearn/pipeline.py#line=620), in Pipeline.fit(self, X, y, **params)
    615     if self._final_estimator != "passthrough":
    616         last_step_params = self._get_metadata_for_step(
    617             step_idx=len(self) - 1,
    618             step_params=routed_params[self.steps[-1][0]],
    619             all_params=params,
    620         )
--> 621         self._final_estimator.fit(Xt, y, **last_step_params["fit"])
    623 return self

File [~/miniconda3/envs/pysindy/lib/python3.14/site-packages/pysindy/optimizers/base.py:176](http://localhost:46683/home/loiseau/miniconda3/envs/pysindy/lib/python3.14/site-packages/pysindy/optimizers/base.py#line=175), in BaseOptimizer.fit(self, x_, y, sample_weight, **reduce_kws)
    173 x_, y = drop_nan_samples(x_, y)
    174 x_, y = check_X_y(x_, y, accept_sparse=[], y_numeric=True, multi_output=True)
--> 176 x, y, X_offset, y_offset, X_scale = _preprocess_data(
    177     x_,
    178     y,
    179     fit_intercept=False,
    180     copy=self.copy_X,
    181     sample_weight=sample_weight,
    182 )
    184 if sample_weight is not None:
    185     x, y = _rescale_data(x, y, sample_weight)

ValueError: too many values to unpack (expected 5, got 6)

PySINDy/Python version information:

• psyndy 2.0.0
• numpy 2.4.0
• scikit-learn 1.8.0

[Jacob-Stevens-Haas]

Good find :) #667 mentioned the same issue - I pushed #670 last night, but I forgot to tag a new release

[Jacob-Stevens-Haas]

2.1.0 now available on pypi