model = LogisticRegression() #algoritmo

my_scorer = make_scorer(score, greater_is_better=True) #Se crea el'score' de la función propia creada
generador_train = GroupKFold(n_splits=10).split(X_train, y_train, order_train) #Generador para creación de los 10 splits siguiendo un orden dado
C= {'C': 10. ** np.arange(-3, 4)} #valores de C
scaler = preprocessing.StandardScaler() #Estandarizado
selector =RFECV(cv=generador_train, estimator=model,scoring=my_scorer #Seleccion de atributos

pipe=Pipeline([('scaler', scaler),('select', selector),('model', model)]) # Se crea la pipeline


grid = GridSearchCV(estimator=pipe, param_grid=C,cv=generador_train,scoring=my_scorer,refit=True) #Se declara el gridSearch con CV

grid.fit(X_train, y_train) # Se ejecuta la pipeline
best_pipe=grid.best_estimator_

can't pickle generator objects