Installing TensorFlow on Pi Supercomputer

Login to CentOS 7 node via servers mu06(202.120.58.230) or mu07(202.120.58.231). Then install pre-built TensorFlow from Anaconda.

$ module load anaconda/2
$ conda create --name deeplearning numpy tensorflow-gpu keras ipython

Activate the tensorflow environment.

$ source activate deeplearning
$ ipython
> import tensorflow as tf

For GPU-TensorFlow jobs, the following lines are required in the job script:

source /usr/share/Modules/init/bash
unset MODULEPATH
module use /lustre/usr/modulefiles/pi
module purge
module load anaconda/2
source activate deeplearning

SLURM partitions centos7gpu, centos7k40, centos7k80 and p100 are ready for these jobs.