Install tensorflow on server

1. Connect to "Robin" which is the Sunnybrook server

ssh <username>@142.76.1.189

2. Connect to lab server "lin1" or "lin2"

ssh lin2

3. You can check the server details:

check the installed softwares:

rpm -qa

check the hardware:

lscpu

4. download anaconda:

wget https://repo.continuum.io/archive/Anaconda3-2018.12-Linux-x86_64.sh

5. check list of folders to ensure it is downloaded:

ls

6. install anaconda:

bash Anaconda3-2018.12-Linux-x86_64.sh

it will ask if you want to install VScode. answer NO

7. check if anaconda installed properly:

conda

8. make your own virtual environment:

conda create --name <name of your environment>

9. activate your env:

conda activate <name of your environment>

10. check the GPU and cuda version of the server:

nvidia-smi

11. if GPU is available install tensorflow-gpu:

conda install tensorflow-gpu

12. check if tensorflow has been installed properly:

python
import sys
import tensorflow as tf
import tensorflow.keras

tf.__version__
tensorflow.keras.__version__

13. check if tensorflow is working properly on the GPU:

tf.config.list_physical_devices('GPU')

above command returns a list of available gpus.

In other word, if len(tf.config.list_physical_devices('GPU'))>0 means that your tf is working on GPU.