I'd like use pearson metric in GridSearch

Question

For Kaggle submission, I need to test my model with pearson indice from scipy. I did this code, but the return is only NaN for mean_train_score and mean_test_score: Who has better idea ?

from scipy.stats import pearsonr
from sklearn.linear_model import LinearRegression

parameters = {}

lr = GridSearchCV(LinearRegression(), 

                  parameters, 
                  scoring=pearsonr, 
                  cv=5, 
                  n_jobs=-1, pearson indice
                  return_train_score=True)
lr.fit(X_train, y_train)

lr.cv_results_