First, you have to sign up for the service. One tip here: students of the fast.ai course get a promo code, which is worth $15. That’s up to about 30 hours of GPU usage, so I suggest you start fast.ai’s excellent course and reap the additional benefits.

Once you signed up, you have access to your console. There are three products available for selection: Gradient°, Core, and API. For fast access to GPU resources, I prefer Gradient°

When you click on Create Notebook +, there is a selection of free containers available at your disposal. You can also configure a new container if necessary. For a quick analysis or experiment, the Jupyter Notebook Data Science Stack is perfect. You can find the specifications

In the second step of creating a notebook instance, you have to decide on a machine you want to use. For example, I use the cheapest CPU option to download data and to do test runs of my deep learning models. This option is called C2, and it costs only $0.009 per hour. That is four and a half days of computing power for $1

Starting with a cheap CPU-solution is possible due to the integrated storage. There are 200GB of permanent storage available, which automatically connect to every new instance via the storage folder. This configuration allows you to prepare everything with a cheap CPU first.

Once you are sure that everything works fine, you can start a new JupyterHub environment. This time, I select a GPU option. The cheapest option starts at $0.51 per hour, and it goes up to $1.72 per hour. There are also more powerful machines available (with a maximum cost of $20.99 per hour), but you need to upgrade your account to do that. I use Paperspace to support my personal development in the area of deep learning. For this purpose, the $0.51 option was always enough.

After you are satisfied with the results, you should export the notebooks you used (both the original file and an HTML-copy) and the trained model to a backup place or version control system of your choice.