Tensorflow with python 3.X version needed

Python 2.7 comes in end of life this January.
Would it be possible to install tensorflow with more recent python version the following months?

What is the procedure to install new software packages on baobab in general?

Thank you in advance.

Hi,
In addition to having new software installed by the admins of baobab, Singularity allows the use of a huge library of software on baobab. More information can be found https://baobabmaster.unige.ch/enduser/src/enduser/applications.html#singularity
Images for tensorflow can be found https://hub.docker.com/r/tensorflow/tensorflow or in the nvidia repository https://ngc.nvidia.com/models .

Using singularity is slightly more difficult but allows greater flexibility in term of available version.

I hope this help.

1 Like

Thank you I will give a try with singularity. By the way, running tensorflow through docker requires usually the use of a separate docker with CUDA drivers. Is there some guide on how to run singularity - cuda setup in baobab?

There is an old documentation here ( Quick guide to Docker, singularity and shifter ).
The nvidia container repository I linked above is setup with all the necessaries drivers. You only need to give the --nv flag to singularity and run on a node with gpus.

1 Like

Dear Dimitirios,

TensorFlow is already installed for Python 3.6.4 on Baobab:

    module spider TensorFlow
TensorFlow:
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
    Description:
      An open-source software library for Machine Intelligence

     Versions:
        TensorFlow/1.7.0-Python-3.6.4
        TensorFlow/1.10.1-Python-2.7.15
1 Like