4

I tried to run tensorflow_gpu on a server with no internet connection.

The script that runs tensorflow_gpu failed with this error:

Intel MKL FATAL ERROR: Cannot load libmkl_avx2.so or libmkl_def.so.

I found a solution to this problem in this post, I have to do the following:

conda install nomkl numpy scipy scikit-learn numexpr
conda remove mkl mkl-service

Since the server has no internet connection, I have completely no idea of how I can install those packages.

What packages do I need to upload to the server in order to mimic conda install?

meTchaikovsky
  • 7,478
  • 2
  • 15
  • 34

1 Answers1

11

This question has been answered before at Stack Overflow: Install anaconda library from a local source

I am copying and rephrasing from the best answer there.

(1) Download the tar files for the desired packages.

Search for packages using a computer that DOES have an internet connection. Copy it over to the server that does NOT have an internet connection. Remember the path.

Example package:
https://repo.continuum.io/pkgs/free/win-64/pymc-2.3.5-np110py35_0.tar.bz2

(2) Then run conda install in offline mode, with server path to tar files. 

conda install --offline $home/savedfile/pymc-2.3.5-np110py35_0.tar.bz2  
# Substitute with your path and directory syntax based on your OS.

This question is old so you may have already found a solution. If not, here it is.

Jennifer Yoon
  • 791
  • 5
  • 10