Behavior of log_loss in scikit learn vis-à-vis numpy arrays

Question

I tried to compute the log loss using log_loss that\'s implemented in scikit learn. First I used

log_loss(y_test,y_pred_NN_2L_ReLu)

And I got