Jupyter Notebook Setup on a VPS

COMPUTER PROJECTS

These steps have been tested on Digital Ocean server and a scaleway ARM64 server. I used the cheapest provided options for my setups.

First, set up ubuntu on the server. Both Digital Ocean and Scaleway provide an easy way to do this. Just follow their prompts, and you'll have your server up and running. After the server is ready, update ubuntu:

sudo apt update
sudo apt upgrade

The server only has the root user. It is not advisable to use root for normal work. So, I create a new user and switch to it:

useradd -m -s /bin/bash user
usermod -aG sudo username
sudo passwd username
su - user

Pip will be used to install various python packages. To install it:

sudo apt install python3-pip

I then install various ml libraries I'll use later:

python3 -m pip install --user numpy scipy matplotlib pandas sympy nose seaborn pillow keras tensorflow scikit-learn scikit-image kaggle-cli
pythom3 -m pip install ipython[notebook] jupyter

When setting up scaleway, some of the libraries failed to install. Here are the dependenices I had to install for pillow:

sudo apt-get install libtiff5-dev libjpeg8-dev zlib1g-dev libfreetype6-dev liblcms2-dev libwebp-dev tcl8.6-dev tk8.6-dev python-tk

And here are the dependencies I had to install for scikit:

sudo apt-get install gcc gfortran python-dev libopenblas-dev liblapack-dev

And for kaggle-cli too, I had to do this:

sudo apt-get install libxml2-dev libxslt1-dev python3-lxml

Tensorflow cannot be installed in the arm64 scaleway instance using pip. So remove it from the packages.

The packages are installed, but cannot be got from the terminal. This is because the bash environment does not have it in its PATH variable. To fix this, add this to the ~/.profile file:

if [ -d "$HOME/.local/bin" ] ; then
        PATH="$HOME/.local/bin:$PATH"
fi

Having set up most of the libraries I'll need, I have to configure jupyter so that I can access it from my browser.

jupyter notebook --generate-config

A file is generated in ~/.jupyter/jupyter_notebook_config.py. This file contains various configuration options for jupyter.

I change the following:

c.NotebookApp.open_browser = False
c.NotebookApp.ip = '0.0.0.0'

To set up the password to be used on logging in do:

jupyter notebook password

I then start jupyter using:

nohup jupyter notebook

The jupyter can be accessed from : ip_address:8888 using any browser.

To install Tensorflow on scaleway's arm processors, I found instructions for a custom build here.

The instructions are to:

curl -L https://github.com/lherman-cs/tensorflow-aarch64/releases/download/r1.4/tensorflow-1.4.0rc0-cp35-cp35m-linux_aarch64.whl > /tmp/tensorflow-1.4.0rc0-cp35-cp35m-linux_aarch64.whl
python3 -m pip install /tmp/tensorflow-1.4.0rc0-cp35-cp35m-linux_aarch64.whl