升级Numpy至1.24后导入TensorFlow遇AttributeError的解决方法

问题原因 #

TensorFlow 2.6.0的底层代码仍在使用np.object这个NumPy别名，但NumPy从1.20版本开始标记该别名废弃，1.24版本直接移除了它，导致导入TensorFlow时触发AttributeError: module 'numpy' has no attribute 'object'。

可行解决办法 #

• 降级NumPy到兼容版本
  TensorFlow 2.6.0官方兼容的NumPy版本上限为1.19.x，执行以下命令降级：

  # pip环境
  pip install numpy==1.19.5
  # conda环境
  conda install numpy=1.19.5
  

• 升级TensorFlow至兼容新版本NumPy的版本
  TensorFlow 2.8及以上版本已修复该兼容性问题，支持NumPy 1.24+。结合Python 3.8的版本支持，可升级到2.10.0等稳定版：

  pip install tensorflow==2.10.0
  

• 临时应急补丁（不推荐长期使用）
  如果暂时无法修改版本，可在导入TensorFlow前手动给NumPy补全object属性：

  import numpy as np
  np.object = object
  import tensorflow as tf
  

  注意：此方法仅为临时 workaround，可能引发其他潜在问题，不建议在生产环境使用。

报错堆栈信息 #

import tensorflow as tf

Traceback (most recent call last):

  Cell In[7], line 1
    import tensorflow as tf

  File ~/opt/anaconda3/envs/profiler/lib/python3.8/site-packages/tensorflow/__init__.py:41
    from tensorflow.python.tools import module_util as _module_util

  File ~/opt/anaconda3/envs/profiler/lib/python3.8/site-packages/tensorflow/python/__init__.py:46
    from tensorflow.python import data

  File ~/opt/anaconda3/envs/profiler/lib/python3.8/site-packages/tensorflow/python/data/__init__.py:25
    from tensorflow.python.data import experimental

  File ~/opt/anaconda3/envs/profiler/lib/python3.8/site-packages/tensorflow/python/data/experimental/__init__.py:97
    from tensorflow.python.data.experimental import service

  File ~/opt/anaconda3/envs/profiler/lib/python3.8/site-packages/tensorflow/python/data/experimental/service/__init__.py:353
    from tensorflow.python.data.experimental.ops.data_service_ops import distribute

  File ~/opt/anaconda3/envs/profiler/lib/python3.8/site-packages/tensorflow/python/data/experimental/ops/data_service_ops.py:26
    from tensorflow.python.data.experimental.ops import compression_ops

  File ~/opt/anaconda3/envs/profiler/lib/python3.8/site-packages/tensorflow/python/data/experimental/ops/compression_ops.py:20
    from tensorflow.python.data.util import structure

  File ~/opt/anaconda3/envs/profiler/lib/python3.8/site-packages/tensorflow/python/data/util/structure.py:26
    from tensorflow.python.data.util import nest

  File ~/opt/anaconda3/envs/profiler/lib/python3.8/site-packages/tensorflow/python/data/util/nest.py:40
    from tensorflow.python.framework import sparse_tensor as _sparse_tensor

  File ~/opt/anaconda3/envs/profiler/lib/python3.8/site-packages/tensorflow/python/framework/sparse_tensor.py:28
    from tensorflow.python.framework import constant_op

  File ~/opt/anaconda3/envs/profiler/lib/python3.8/site-packages/tensorflow/python/framework/constant_op.py:29
    from tensorflow.python.eager import execute

  File ~/opt/anaconda3/envs/profiler/lib/python3.8/site-packages/tensorflow/python/eager/execute.py:27
    from tensorflow.python.framework import dtypes

  File ~/opt/anaconda3/envs/profiler/lib/python3.8/site-packages/tensorflow/python/framework/dtypes.py:585
    np.object,

  File ~/opt/anaconda3/envs/profiler/lib/python3.8/site-packages/numpy/__init__.py:305 in __getattr__
    raise AttributeError(__former_attrs__[attr])

AttributeError: module 'numpy' has no attribute 'object'.
`np.object` was a deprecated alias for the builtin `object`. To avoid this error in existing code, use `object` by itself. Doing this will not modify any behavior and is safe. 
The aliases was originally deprecated in NumPy 1.20; for more details and guidance see the original release note at:
    https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations

内容的提问来源于stack exchange，提问作者user62198