[警告] FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future ver

警告信息

FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.
  _np_qint8 = np.dtype([("qint8", np.int8, 1)])
FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.
  _np_quint8 = np.dtype([("quint8", np.uint8, 1)])
FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.
  _np_qint16 = np.dtype([("qint16", np.int16, 1)])
FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.
  _np_quint16 = np.dtype([("quint16", np.uint16, 1)])
FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.
  _np_qint32 = np.dtype([("qint32", np.int32, 1)])
FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.
  np_resource = np.dtype([("resource", np.ubyte, 1)])
FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.
  _np_qint8 = np.dtype([("qint8", np.int8, 1)])
FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.
  _np_quint8 = np.dtype([("quint8", np.uint8, 1)])
FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.
  _np_qint16 = np.dtype([("qint16", np.int16, 1)])
FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.
  _np_quint16 = np.dtype([("quint16", np.uint16, 1)])
FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.
  _np_qint32 = np.dtype([("qint32", np.int32, 1)])
FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.
  np_resource = np.dtype([("resource", np.ubyte, 1)])

系统环境

• Ubuntu 18.4 LTS
• Python 3.6
• TensorFlow 2.0.0-beta1
• NumPy 1.17.2

原因分析

在TensorFlow 2.0.b1的版本下，如果Numpy的版本超过了1.17，那么就会出现这个警告。
这个问题已经在 GitHub 的 PR #30559 [2] 中得到了解决，不出意外的话，在TensorFlow的下个版本中，应该就不会再出现这个警告了。

解决方案

有如下三种临时解决方案：

1. 回退版本

既然高版本（Numpy 1.17.0rc1)会报错，那么稍微降低版本不就可以了！？！

pip install "numpy<1.17"

2. 暂时忽略报错

根据网友 njanirudh 和 Gra-tak 的解决方案 [1]，暂时忽略报错是一个不错的选择，毕竟这个报错目前还没有负面影响。在导入tensorflow之前，导入warnings包：

import warnings
warnings.filterwarnings('ignore', category=FutureWarning)

import tensorflow as tf

当然，这会暂时忽略所有的FutureWarning报错。

3. 物我两忘

它报任它报，物我而兼忘。

更新

实测TensorFlow 2.0.0-rc1不会报警告。

Reference

1. njanirudh, Gra-tak. (September 22, 2019). FutureWarning: Deprecated numpy API calls in tf.python.framework.dtypes. Retrieved from https://github.com/tensorflow/tensorflow/issues/30427
2. yongtang. (September 22, 2019). Fix numpy warning with numpy 1.17.0+. Retrieved from https://github.com/tensorflow/tensorflow/pull/30559