You want to start using Tensorflow, but you heard that training time and other main tasks related to Deep Learning take forever on simple desktops?
Deep learning is one of those exciting (almost) new things that are changing almost every industry we know. Learning to use tools such as Tensorflow, PyTorch and Keras is super simple today.
There are tutorials, free courses, books and a lot of other resources that will help you to acquire this new skill very quickly.
The problem is… equipment.
Some of the fundamental building blocks of deep learning can take forever on an ordinary desktop.
Preprocessing data is already challenging when we are talking about Big Data. But training very large sets of data is nearly impossible at home.
So, how to solve this contradiction? You can learn, but you can’t practise, right?
Not exactly. There are some tools like Colab Notebooks that mimic the famous Jupyter Notebooks online. In this way, you can use Google infrastructure to practise, prototype and learn fundamental concepts before embarking on a real-world project.
How Colab works
Starting a Colab is as simple as connecting to your Google account and clicking on this link.
Doing so will create your first online notebook. If you already know Jupyter Notebooks, keep reading this text.
If you don’t, get familiar with Jupyter here. Come back after you have played a little with some notebooks on your computer 🙂
When you create your first online Colab, you will see this screen:
Colab is describe as “a Python development environment that runs in the browser using Google Cloud”.
Some of the coolest features of this method is the possibility of saving your work as a file on your Google Drive or on GitHub (you can also save it as a Gist).
You can even download it if you need for later use on your desktop.
Some things to care about
At some point working with Colab, you may want to use a library you have developed. You can install new libraries using pip or apt-get.
!pip install -q matplotlib-venn !apt-get -qq install -y libfluidsynth1
It is also possible to connect to a local or hosted execution environment.
These are some of the basic Colab features. If you wish, Udacity has this great video showing how Colab behaves and how to make your first steps on the platform.
Do you want to connect? It will be a pleasure to discuss Machine Learning with you. Drop me a message on LinkedIn.