Hardware and software setup - basic python and jupyter notebooks

In a Nutshell

If you are completely new to the whole python/deep-learning environment, here are a few facts you need to know.

  • python is a programming language that can run on any OS and is, de facto, becoming the most used language in deep-learning.

  • a notebook is an interface to an IPython kernel, a background process that execute python code. In a notebook, you can store, edit and execute python code and interact with whatever this code outputs.

  • mimikit is a package that contains the python code that is then invoked or called by the mimikit-notebooks.

Hardware Setup

Doing deep-learning requires a GPU (Graphics Processing Unit), simply because deep-learning algorithms rely on certain computations for which GPUs have been designed. Although running those algorithms on a CPU is perfectly fine, it is, unfortunately, prohibitively slow.

Furthermore, deep-learning libraries specifically need NVIDIA GPU and are not compatible with models from other manufacturer.

GPU Providers

If you don’t own a NVIDIA GPU, there exists several online GPU provider that offer limited free access along with various paid plans. For instance :

When using those services, you essentially rent or borrow a computer with a GPU and connect to it through a web page in your browser (a notebook).

If you have access to a NVIDIA GPU, you only need to make sure that an appropriate driver is installed on your system, before you go through the Software installation steps.

Software Installation

On your system

In order to run the mimikit-notebooks on your system, you’ll need Linux or MacOS with python>=3.8 and pip installed. Once you do, open a terminal and enter the command-line

pip install mimikit

This will download mimikit and install it on your system.

  • create a directory in your User/home/, something like User/home/mmk-stuff/.

  • under mmk-stuff/, create two directories data/ and notebooks/.

  • open a browser and go to the list of plain notebooks on github.

  • Click on the one you want to download on the page showing its code

  • click on the Raw button on the right.

  • Hit Cmd+S or Ctrl+S to save the notebook on your system in mmk-stuff/notebooks

  • Now, open a terminal and cd to mmk-stuff

  • Finally, enter the command-line

jupyter notebook

This will launch an IPython kernel and open a new window in your browser. In this window, navigate to where you downloaded a notebook and click on it.

With a GPU Provider

Because GPU providers want to make your life easier, they all offer kernels with software commonly used for deep-learning pre-installed. Every time you open a kernel or a session, you get a “new computer” with python and the necessary deep-learning libraries already installed.

The only piece of software you’ll still need to install every time you open a session is mimikit. To do so, one only needs to execute a code cell with the following line

!pip install mimikit

Please note that all the colab mimikit-notebooks already have this cell so that running a notebook always includes installing mimikit.

Basic Python - Jupyter Notebook

Although running the mimikit-notebooks doesn’t require you to be able to write python code, you might find it useful to know the basics. Here, we simply point to resources we find useful and invite you to check them out.