Hmm. I am checking dtypes via `np.floating` and `np.signedinteger`. What will this change to?