无法导入 conda 环境中安装的 package

Question

There are similar previous questions, eg Can't import package after installing in a conda environment or ModuleNotFoundError when importing package that is installed in conda environment , but I couldn't find a solution that worked in any of the existing questions.

I have a conda env called 'keras', and I installed various packages, including keras in it. Eg if I try to install Keras again I get:

C:\Users\Ori Family>conda activate keras

(keras) C:\Users\Ori Family>conda install -c conda-forge keras
Collecting package metadata (current_repodata.json): done
Solving environment: done

# All requested packages already installed.

But the module is not actually available. Eg:

(keras) C:\Users\Ori Family>python
Python 3.9.7 (default, Sep 16 2021, 16:59:28) [MSC v.1916 64 bit (AMD64)] :: Anaconda, Inc. on win32
Type "help", "copyright", "credits" or "license" for more information.
>>> from keras.models import Sequential
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ModuleNotFoundError: No module named 'keras'

Everything seems to be in order:

>>> import platform
>>> print(platform.architecture())
('64bit', 'WindowsPE')
>>> import sys
>>> for p in sys.path:
...   print(p)
...

C:\tools\anaconda3\envs\keras\python39.zip
C:\tools\anaconda3\envs\keras\DLLs
C:\tools\anaconda3\envs\keras\lib
C:\tools\anaconda3\envs\keras
C:\Users\Ori Family\AppData\Roaming\Python\Python39\site-packages
C:\tools\anaconda3\envs\keras\lib\site-packages
>>> print(sys.executable)
C:\tools\anaconda3\envs\keras\python.exe

I'm using conda 4.11.0, and conda list -n keras has this line in the output:

keras                     2.6.0            py39hd3eb1b0_0

Any tips on how to solve/diagnose/debug this?

Answer 1

You need to install Tensorflow and to change the way you import Keras -> tensorflow.keras.models .

# create a fresh env for keras
conda create -c conda-forge -n keras python=3.7 keras tensorflow
# use the env
conda activate keras

Then you can import it.

import tensorflow as tf
from tensorflow import keras
from tensorflow.keras.models import Sequential
print(f"{keras.__version__}")
# 2.6.0