Memory error SparseLinearDML

Try to make a model with SparseLinearDML but get a memory error:

import numpy as np
from sklearn.preprocessing import PolynomialFeatures
from econml.dml import SparseLinearDML
from catboost import CatBoostClassifier, CatBoostRegressor

est = SparseLinearDML(model_y=CatBoostClassifier(max_depth=4,
                                                 n_estimators=10,
                                                 verbose=False
                                                ),
                      model_t=CatBoostRegressor(max_depth=4,
                                                 n_estimators=10,
                                                 verbose=False
                                                ),
                      linear_first_stages=False,
                      n_splits=2,
                      random_state=1,
                      discrete_treatment=False
                    )
N=5000000
cols = 150
train = np.random.random(size=(N, cols))
target = np.random.choice([0, 1], size=N, replace=True)
treatment =  np.random.choice([0, 100, 500, 1000, 5000], size=N, replace=True)
est.fit(target, treatment, X=train)

KernelNotResponding: Kernel died unexpectedly and has been restarted. If it's not coming back, please, try restarting from the main menu.

Working on 64 kernel threads and 500gb
Is there a limit on row or column number? Or maybe catboost is not a good choice for T and Y models..? How can I solve this problem?

Start by reproducing the reported SparseLinearDML.fit call with the supplied dataset dimensions and CatBoost models. Inspect the SparseLinearDML entry point and determine whether the kernel failure comes from a documented row or column limit, model configuration, or memory use during fitting. Done means identifying a reproducible cause and documenting or resolving the failure.