属性 AttributeError：模块“tensorflow.compat.v2”没有属性“_internal_”

Question

按照 CARLA 教程，我创建了一个文件来训练我的模型； 但是，每当我尝试使用命令提示符运行它时，都会出现此错误：

2021-06-20 19:39:30.429984: W tensorflow/stream_executor/platform/default/dso_loader.cc:59] Could not load dynamic library 'cudart64_101.dll'; dlerror: cudart64_101.dll not found
2021-06-20 19:39:30.430242: I tensorflow/stream_executor/cuda/cudart_stub.cc:29] Ignore above cudart dlerror if you do not have a GPU set up on your machine.
Using TensorFlow backend.
Traceback (most recent call last):
  File "CARtut5.py", line 10, in <module>
    from keras.applications.xception import Xception
  File "C:\Users\user\AppData\Local\Programs\Python\Python37\lib\site-packages\keras\_init_.py", line 12, in <module>
    from . import initializers
  File "C:\Users\user\AppData\Local\Programs\Python\Python37\lib\site-packages\keras\initializers\_init_.py", line 124, in <module>
    populate_deserializable_objects()
  File "C:\Users\user\AppData\Local\Programs\Python\Python37\lib\site-packages\keras\initializers\_init_.py", line 49, in populate_deserializable_objects
    LOCAL.GENERATED_WITH_V2 = tf._internal_.tf2.enabled()
AttributeError: module 'tensorflow.compat.v2' has no attribute '_internal_'

我尝试卸载Keras和TensorFlow并重新安装它们，但错误仍然存​​在。

Answer 1

来自评论

尝试恢复到旧版本的tf ，这可能会有所帮助（转述自 StarShine）

其实上述问题是由于最新之间的不兼容Tensorflow版本和独立Keras 。 有关更多信息，您可以参考此处。