 I already updated numpy to the latest version. Is there any other ways to Save bfloat16 as binary format?