TensorFlow Keras导入报错：重复注册ReduceDataset梯度

问题场景 #

使用Keras构建简单分类网络，导入必要包时触发错误。

导入代码 #

from tensorflow import keras
from keras.models import Sequential
from keras.layers import Dense
from keras.callbacks import EarlyStopping, ModelCheckpoint
from scikeras.wrappers import KerasClassifier

报错信息 #

KeyError: "Registering two gradient with name 'ReduceDataset'! (Previous registration was in register /Library/Frameworks/Python.framework/Versions/3.12/lib/python3.12/site-packages/tensorflow/python/framework/registry.py:65)"

报错栈起始信息 #

---------------------------------------------------------------------------
KeyError                                  Traceback (most recent call last)
Cell In[126], line 1
----> 1 from tensorflow import keras
      3 from keras.models import Sequential
      4 from keras.layers import Dense

File /Library/Frameworks/Python.framework/Versions/3.12/lib/python3.12/site-packages/tensorflow/__init__.py:45
     42 from tensorflow.python import tf2 as _tf2
     43 _tf2.enable()
----> 45 from tensorflow._api.v2 import __internal__
     46 from tensorflow._api.v2 import __operators__
     47 from tensorflow._api.v2 import audio

File /Library/Frameworks/Python.framework/Versions/3.12/lib/python3.12/site-packages/tensorflow/_api.v2/__internal__/__init__.py:11
      9 from tensorflow._api.v2.__internal__ import decorator
     10 from tensorflow._api.v2.__internal__ import dispatch
----> 11 from tensorflow._api.v2.__internal__ import distribute
     12 from tensorflow._api.v2.__internal__ import eager_context
     13 from tensorflow._api.v2.__internal__ import feature_column

...

环境信息 #

• 系统：macOS
• 已安装包版本：

tensorflow 2.16.1
keras 3.3.3
scikeras 0.13.0
scikit-learn 1.4.2

解决方案 #

1. 统一Keras导入方式
   Keras 3.x是独立多后端框架，混合导入tensorflow.keras和独立keras会引发冲突，二选一即可：

• 仅使用独立Keras导入：

  from keras.models import Sequential
  from keras.layers import Dense
  from keras.callbacks import EarlyStopping, ModelCheckpoint
  from scikeras.wrappers import KerasClassifier
  

• 或统一使用TensorFlow后端的Keras路径：

  from tensorflow.keras.models import Sequential
  from tensorflow.keras.layers import Dense
  from tensorflow.keras.callbacks import EarlyStopping, ModelCheckpoint
  from scikeras.wrappers import KerasClassifier
  

2. 升级Scikeras适配Keras 3.x
   Scikeras 0.13.0对Keras 3.x的支持存在适配问题，升级到兼容版本：

pip install --upgrade scikeras

3. 重装匹配版本组合
   若升级后仍报错，卸载现有依赖后重新安装兼容版本：

pip uninstall tensorflow keras scikeras -y
pip install tensorflow==2.16.1 keras==3.3.3 scikeras==0.14.0

内容的提问来源于stack exchange，提问作者all.m