cols = df.select_dtypes([np.number]).columns

Read more of this post