GPU server setup

Setting up the Kernal

Installing mainline

sudo add-apt-repository ppa:cappelikan/ppa
sudo apt update
sudo apt install mainline -y
sudo mainline install 6.2

Reboot the Server

sudo reboot

Remove any existing Nvidia installations

sudo apt install gcc
sudo apt-get remove --purge '^nvidia-.*'
sudo apt-get autoremove

Disabling the default NVidia libraries

Open the file

sudo vi /etc/modprobe.d/blacklist-nvidia-nouveau.conf

Insert these lines

blacklist nouveau
options nouveau modeset=0

Save and exit. Regenerate the kernel initramfs.

sudo update-initramfs -u

Check the graphics card by executing the following command:

sudo lshw -C display

Your graphic cards should be shown in the output generated.

Installing Nvidia CUDA libraries

Visit the following URL: https://developer.nvidia.com/cuda-12-2-0-download-archive
Choose your configurations based on your Operating System. Please Choose X86 architecture.

Nvidia will provide you a set of commands to execute. Please execute them one after another by copying and pasting them. An example has been provided below:

wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/cuda-ubuntu2204.pin
sudo mv cuda-ubuntu2204.pin /etc/apt/preferences.d/cuda-repository-pin-600
wget https://developer.download.nvidia.com/compute/cuda/12.2.0/local_installers/cuda-repo-ubuntu2204-12-2-local_12.2.0-535.54.03-1_amd64.deb
sudo dpkg -i cuda-repo-ubuntu2204-12-2-local_12.2.0-535.54.03-1_amd64.deb 
sudo cp /var/cuda-repo-ubuntu2204-12-2-local/cuda-*-keyring.gpg /usr/share/keyrings/
sudo apt update && sudo apt -y install cuda 

sudo reboot

After rebooting the server, execute the following command:
```
nvidia-smi
```
You should now see the list of GPUs attached to your machine.

Installing Nvidia cuDNN

Using your Nvidia account, visit the following URL: https://developer.nvidia.com/rdp/cudnn-archive
Download the local installer by choosing the latest version based on your operating system.

Execute the following commands:

wget -nc https://www.dropbox.com/scl/fi/ck7akyej5l0onopvmvecf/cudnn-local-repo-ubuntu2204-8.9.7.29_1.0-1_amd64.deb

sudo dpkg -i cudnn-local-repo-ubuntu2204-8.9.7.29_1.0-1_amd64.deb
sudo cp /var/cudnn-local-repo-ubuntu2204-8.9.7.29/cudnn-local-08A7D361-keyring.gpg /usr/share/keyrings/
sudo apt update && sudo apt install libcudnn8 libcudnn8-dev -y 

sudo useradd -m -s /bin/bash -G sudo jupyter

sudo passwd jupyter

su - jupyter

Setting a Virtual Environment using Conda

Download the latest version of Anaconda.

wget -nc https://repo.anaconda.com/archive/Anaconda3-2024.10-1-Linux-x86_64.sh

Install the conda

bash Anaconda3-2024.10-1-Linux-x86_64.sh

Press Enter
Press Spacebar until you see license terms
Type Yes and press Enter
Press Enter
Type Yes to install the path variable.

If you face any problem, type the following command:
```
  eval "$(/home/jupyter/anaconda3/bin/conda shell.bash hook)"
```

Execute the below provided commands

source .bashrc
conda config --set auto_activate_base false 

Check the Python Version
```
python -V
```
Tensorflow GPU currently works better if your Python Version is 3.10. Thus, let us create virtual environment with Python 3.10
```
conda create --name jupyterHub python=3.10 -y
```
Enter into that Virtual Environment
```
conda activate jupyterHub
```

Installing Tensorflow, TensorRT, and other import DL-plugins

Add the necessary channels

pip install --upgrade pip
conda config --add channels conda-forge
conda config --add channels microsoft

Install tensorflow-gpu

conda install -n base conda-libmamba-solver -y
conda config --set solver libmamba
conda install anaconda::numpy -y
conda install anaconda::scipy -y
conda install conda-forge::cupy -y
conda install anaconda::tensorflow-gpu -y

Check the correctness of installed TensorFlow-gpu by executing the following commands:

python3 -c "import tensorflow as tf; print(tf.config.list_physical_devices('GPU'))"

Output must contain the following last line

[PhysicalDevice(name='/physical_device:GPU:0', device_type='GPU'), PhysicalDevice(name='/physical_device:GPU:1', device_type='GPU')]

If you do not see the above line, it means their may be an installation problem. Try to execute the below step

 pip install tensorflow[and-cuda]
   
 python3 -c "import tensorflow as tf; print(tf.config.list_physical_devices('GPU'))"

Installation of JupyterHub, JupyterLab, and Notebook

pip install --upgrade pip
conda install -c conda-forge jupyterlab jupyterhub -y
conda install notebook -y

Installing important dependencies, kernels, and widgets of JupyterHub

conda install sidecar playwright -y

playwright install-deps
playwright install

conda install -c conda-forge jupyterlab_widgets -y
conda install conda-forge::configurable-http-proxy -y
conda install -c rapidsai-nightly -c conda-forge jupyterlab-nvdashboard -y
conda install -c conda-forge ipywidgets -y
conda install -c conda-forge jupyterlab-spellchecker -y
conda install conda-forge::jupyterlab-spreadsheet-editor -y
conda install -c conda-forge jupyterlab jupyterlab-git -y

conda install -c conda-forge python-kaleido -y
conda install conda-forge::jupyterlab-latex -y
conda install -c conda-forge jupyterlab-drawio -y

conda install bokeh -y
conda install -c plotly plotly=6.0.0rc0 -y
conda install -c plotly plotly-geo=1.0.0 -y

conda install -c conda-forge jupyter_contrib_nbextensions

sudo apt install -y qtcreator qtbase5-dev qt5-qmake cmake

conda install qtconsole -y

pip install pyqt5

pip install aquirdturtle_collapsible_headings 

pip install tqdm jupyterlab-unfold
pip install jupyterlab-code-formatter 
pip install black isort
pip install lckr_jupyterlab_variableinspector
 

pip install jupyterlab_theme_solarized_dark

pip install pyqt5
conda install qtconsole -y

pip install colabcode
pip install autots auto-ts darts


#pip install autots auto-ts darts etna[all] greykite #kats
#error in kats setupfile. 

jupyter contrib nbextension install --user

Setting up the JupyterHub

Come to the home directory.

cd ~

Setup symbolic links to jupyterHub and jupyterNotebook to access them directly from the terminal.

sudo ln -s ~/anaconda3/envs/jupyterHub/bin/jupyter /usr/bin/jupyter
sudo ln -s ~/anaconda3/envs/jupyterHub/bin/jupyterhub /usr/bin/jupyterhub

Create the jupyterHub default configuration file

jupyterhub --generate-config -f ~/jupyterhub_config.py

Open the jupyterHub configuration file

vi ~/jupyterhub_config.py

Add the following text in the file

c.Authenticator.allow_all = True
c.PAMAuthenticator.open_sessions = False
c.JupyterHub.bind_url = 'http://163.143.87.224:80'   #REPLACE the IPADDRESS
c.Spawner.cmd = ['/home/jupyter/anaconda3/envs/jupyterHub/bin/jupyterhub-singleuser']
c.JupyterHub.hub_bind_url = 'http://127.0.0.1:8085'
c.JupyterHub.hub_port = 8082
c.ConfigurableHTTPProxy.command = '/home/jupyter/anaconda3/envs/jupyterHub/bin/configurable-http-proxy'  #'/usr/local/bin/configurable-http-proxy'
c.Spawner.cmd=["/home/jupyter/anaconda3/envs/jupyterHub/bin/jupyter-labhub"]

#c.JupyterHub.ssl_key = '/path/to/my.key'
#c.JupyterHub.ssl_cert = '/path/to/my.cert'

Note: Change the ipaddress of the variable c.JupyterHub.bind_url

Create a startup file that starts jupyterHub right after the server starts

sudo vi /usr/lib/systemd/system/jupyterhub.service

Copy and Paste the below-provided text:

[Unit]
Description=JupyterHub

[Service]
Type=simple
PIDFile=/run/jupyterhub.pid
# Step 1 and Step 2 details are here..
# ------------------------------------
ExecStart=jupyterhub -f /home/jupyter/jupyterhub_config.py
#User=root
#Group=root
#WorkingDirectory=/home/jupter
Restart=always
RestartSec=10
#KillMode=mixed

[Install]
WantedBy=multi-user.target

Save and exit.

Execute the below commands on the terminal

sudo systemctl enable jupyterhub.service
sudo systemctl daemon-reload
sudo systemctl restart jupyterhub.service

Change the home directory permissions

sudo chmod -R 755 .

Reboot the system

sudo reboot

Setting up the cluster

pip install ipyparallel
ipcluster start