Get hands on - use shared conda installation¶

To use the shared installation you have to tell linux where it is.

  1. Log in ssh ...@gate.ikw.uos.de

  2. Add some lines at the end of your bashrc file:
    $ nano -Bu ~/.bashrc
    Or you could use ranger or vim instead…

    # >>> conda initialize >>>
# !! Contents within this block are managed by 'conda init' !!
__conda_setup="$('/net/projects/scratch/winter/valid_until_31_July_2020/asparagus/sharedConda/bin/conda' 'shell.bash' '$if [ $? -eq 0 ]; then
    eval "$__conda_setup"
else
    if [ -f "/net/projects/scratch/winter/valid_until_31_July_2020/asparagus/sharedConda/etc/profile.d/conda.sh" ]; then        . "/net/projects/scratch/winter/        valid_until_31_July_2020/asparagus/sharedConda/etc/profile.d/conda.sh"
    else
    export PATH="/net/projects/scratch/winter/valid_until_31_July_2020/asparagus/sharedConda/bin:$PATH"
    fi
fi
unset __conda_setup
# <<< conda initialize <<<

  3. For changes to take effect, close and re-open your current shell.
    $ exit

  4. Log in again. If you want to use a single GPU log in to shadow or light. You can test it with: $ nvidia-smi

  5. If you prefer that condas base environment is not activated on startup,
    set the auto_activate_base parameter to false:
    $ conda config --set auto_activate_base false

  6. After Log in hit:

    $ conda activate dataSet or $ conda activate /net/projects/scratch/winter/valid_until_31_July_2020/asparagus/sharedConda/envs/dataSet/

  7. Test a tensorflow file on the GPU
    $ nvidia-smi -l

Check this link out, if you want to know more about shared installations: https://docs.anaconda.com/anaconda/install/multi-user/