thunder_tensorflow_gpu_conda.md

ssh thunder, GPU, tensorflow, jupyter

Conda cheat sheet: https://kapeli.com/cheat_sheets/Conda.docset/Contents/Resources/Documents/index

You should be logged in to OU's VPN to connect to thunder. Use the following link for info about setting up VPN. https://www.ou.edu/marcomm/cms/get-started/vpn

Steps:

(1) Install Anaconda
(2) Create and activate a new conda env with tensorflow-gpu
(3) Install jupyter lab and other packages
(4) Run jupyter lab

(1) Retrieve and install Anaconda from home directory.

cd ~
wget https://repo.anaconda.com/archive/Anaconda3-2021.05-Linux-x86_64.sh
bash Anaconda3-2021.05-Linux-x86_64.sh

(2) Run the bashrc file after installing anaconda.

. ~/.bashrc

After executing the .bashrc file, the (base) environment will become active.

Deactivate the (base) environment before creating new environment.

conda deactivate

(2) Upload the .yaml file to thunder (remember the file location for next command).

Create new conda environment using .yaml file.

conda env create -f ~/conda-yaml/tf-gpu9.yaml

Activate the new environment.

conda activate tf-gpu9

Confirm GPU access in conda environment with tensorflow.

python -c 'import tensorflow as tf ; tf.config.list_physical_devices()'

*** all steps below should have environment activated ***

(3) Install jupyter lab in the environment.

conda install jupyterlab

(4) Launch jupyter lab.

jupyter lab

Create ssh-tunnel from local command line.

Or upload and run script jupy.sh.

ssh -N -f -L 8888:localhost:8888 jreynolds@thunder.cs.nor.ou.edu
-- enter password

Copy and paste the link given from the output of launching jupyter lab from thunder.

http://localhost:8888/lab?token=381df2b315147a55967b758163031eb59e06103929ae704d

Unleash the thunder.