make the expected numpy code of round_half_away_from_zero output the good dtype when dealing with float32.
拖放文件到此处或者 点击上传
