当父目录已经有该库的旧版本时，Python 从另一个目录导入更新的库

Question

我正在开发一个运行 Red Hat Enterprise Linus - 6+ 的强大集群。

我需要从另一个目录导入一个库（比如 Tensorflow），但主目录已经包含旧版本的 tensorflow。

结构 -

I Don't have Admin privileges for these libraries

Anaconda-directory 
 |--- tensorflow 0.7
 |--- other libraries which I need

Other Directory
 |--- tensorflow 0.11

My Home Directory (I have privileges here)
Local directory of anaconda
 |--- Keras

所以，我需要使用 Anaconda 目录，但我想使用 tensorflow (0.11) 的更新版本，我也想使用 Keras。

我尝试了以下命令 -

METHOD - 1
# Load Anaconda module
module load apps/anaconda/4.1.1/gnu
# Has tensorflow 0.7 installed
jupyter notebook
--> Works, but with tensorflow 0.7 (Which is incompatible with keras)

METHOD - 2
# Load new version of tensorflow
# It automatically python module located at compiler/python/2.7.10/compilervars along with CUDAnn and others
module load apps/tensorflow/0.11/gnu

# Point to my local installation of anaconda where I have installed Keras
export PYTHONPATH=/home/-path-to-my-home-dir/anaconda2/lib/python2.7/site-packages/:$PYTHONPATH

echo $PYTHONPATH
/home/-path-to-my-home-dir/anaconda2/lib/python2.7/site-packages/:/home/apps/TENSOR_GPU/0.11/gnu/lib/python2.7/site-packages/

--> Works well in python shell, but I want to use IPython Notebook

jupyter notebook
--> Doesn't work in Ipython Notebook, which I guess is due to the reason that it uses jupyter notebook from my home dir (Not an Admin)


###  WHAT I WANT  ###
Anaconda Distribution --> From Server Module Directory (Not my local Directory)
Jypyter --> From above anaconda dist (Not my local anaconda dist)
Tensorflow --> From another module (Newer Version)
Keras --> Installed in my local anaconda distribution

# Any Other additional packages
# either not available on server (I can't install on server)
# or older versions available
# For such libraries, I should be able to use just that package 
# from my home directory (I can have anaconda installed on my home dir)

Answer 1

我会试试这个。

• 打开一个 Jupyter 笔记本。
• import sys
• sys.path[0:0] = ["/path/to/new/tensorflow"]
• 导入 tensorflow，检查版本。
• 从 Anaconda 导入库。

第三步可以根据您的喜好进行扩展，包括与您的正常PYTHONPATH 不同的其他内容。