8

Following error shows when I import a package, class or method from tensorflow:

Cannot find reference 'keras' in '__init__.py' less... (Ctrl+F1) 
Inspection info: This inspection detects names that should resolve but don't. Due to dynamic dispatch and duck typing, this is possible in a limited but useful number of cases. Top-level and class-level items are supported better than instance items.

This issue did not happen when using preview versions of tensorflow. However, the program still runs normally. I find package 'keras' in 'tensorflow.keras' with this error occurs in the coding interface, but it is unable to run when I change to import 'tensorflow.python.keras'. How do I solve this problem?

I am using tensorflow 2.0.0alpha and lastest version of Pycharm

PyCharm 2018.3.5 (Professional Edition)
Build #PY-183.5912.18, built on February 26, 2019
Licensed to xxx
Subscription is active until May 14, 2019
For educational use only.
JRE: 1.8.0_152-release-1343-b28 amd64
JVM: OpenJDK 64-Bit Server VM by JetBrains s.r.o
Windows 10 10.0
vzZ
  • 89
  • 1
  • 2

1 Answers1

1

I am currently using the TensorFlow 2 beta version installed and used to get red squiggles for the packages I imported from keras (eg. when using from tensorflow.keras import datasets, layers, models Pycharm could not make sense of datasets, layers or models).

I found that manually importing the keras package got rid of the error highlights. Go to Project Interpreter (cmd + , on macOS), click the + button, in the search bar type in "keras", choose the version you need and click on the "Install Package" button.

Madhav M K
  • 77
  • 8
cadebe
  • 651
  • 1
  • 12
  • 35