Palmetto deep learning environment configuration

Get a compute node:

qsub -I -l select=1:ncpus=4:ngpus=1:gpu_model=k40:mem=16gb,walltime=10:00:00

For Python 2.7

Installation:

# Add modules
module add anaconda/4.3.0
module add cuda-toolkit/8.0.44
module add cuDNN/8.0v6

# Create anaconda environment profile and activate it
conda create -n tfpy2
source activate tfpy2

# Check python version and packages:
# python -V
# pip list

# Install tensorflow-gpu locally
pip install --ignore-installed --upgrade --user tensorflow-gpu==1.4.1

# Finished!
# If you want to deactivate tfpy2, just
# source deactivate tfpy2

For next time using:

module add anaconda/4.3.0
module add cuda-toolkit/8.0.44
module add cuDNN/8.0v6
source activate tfpy2
python
...

For Python 3.4

Installation:

# Add modules
module add anaconda3/4.3.0
module add cuda-toolkit/8.0.44
module add cuDNN/8.0v6

# Create anaconda environment profile and activate it
conda create -n tfpy3
source activate tfpy3

# Install tensorflow-gpu locally
pip install --ignore-installed --upgrade --user tensorflow-gpu==1.4.1

# Finished!
# If you want to deactivate tfpy3, just
# source deactivate tfpy3

For next time using:

module add anaconda3/4.3.0
module add cuda-toolkit/8.0.44
module add cuDNN/8.0v6
source activate tfpy3
python
...

Testing Code:

import tensorflow as tf
hello = tf.constant('Hello, TensorFlow!')
sess = tf.Session()
print(sess.run(hello))