Paperspace Launches New Cloud IDE

They join Colab and Deepnote with a custom Jupyter notebook UI

Robert Ritz
Towards Data Science

--

The new IDE in dark mode

Note: This release seems to be very much a work in progress. Check out the comparison to Colab and Deepnote at the bottom for details.

On February 11, 2020, Paperspace rolled out their reimagined cloud-based notebook product with a new IDE and file browser. Paperspace has offered one-click Jupyter notebooks since 2018 through its Gradient product. What has changed is that the default view for notebooks is now their own custom IDE user interface, similar to what Google Colab and Deepnote offer.

Before this update, you chose your base environment with a list of pre-installed packages and a machine type (from free to $$$$). Then you launched a browser tab with Jupyter Notebook or Jupyter Lab. Now you still have the option to use the vanilla Jupyter Notebook/Lab UI, but you can also use a custom notebook user interface that Paperspace claims is much faster and more responsive.

Let’s take a look at the new features in this recent update.

New IDE

Since 2018 Gradient notebooks used the standard Jupyter Notebook or Jupyter Lab UI. This usually meant several minutes to start your environment container, and then you had to load the UI in your browser. Once it was all loaded, it was quite snappy, but that initial loading time was always a barrier to getting my notebooks going fast.

The new IDE allows you to browse files and open notebooks in read-only mode without having a connected machine instance to solve this issue. This means you can get started viewing your notebooks quickly.

The new UI in dark mode.

Of course, you also have a light mode available if dark mode isn’t your thing.

When you are ready, you can choose your machine type and start your instance. Then you will be able to edit and execute your notebooks. The claim on the announcement of a significantly faster setup didn’t seem to happen for me. I had the usual 5–10 minute way to have my instance ready to go. Hopefully, these are launch glitches, and the speed will improve.

Once your instance is ready to go, you now have nice hardware indicators to let you know your CPU and RAM usage.

Unfortunately, in testing, not all environments seem to be compatible with this new IDE. Some environments I tested gave me an error that the notebook is not supported by Gradient yet, but I still have the option to open in classic Jupyter.

Coming Soon…Storage Improvements?

At the bottom of the Paperspace blog announcement, they tease new storage capabilities that will be coming soon. They may be working on a way to solve one of the most significant issues data scientists face today, file and data management. Data sets may be in an S3 bucket or similar object storage, while notebooks may in on a local hard drive, Google Drive, or another cloud provider. This creates a lot of basic time consuming administrative work that is not fun for anyone.

Comparison to Deepnote and Colab

Google Colab offers a free notebook environment that works with Google Drive and gives free GPU access. Deepnote took that a step further and offered Google Doc like functionality in a notebook, where multiple collaborators can code together on the same document in real-time. Deepnote currently doesn’t offer GPU access.

Currently, the updated Gradient Notebooks don’t offer as much functionality as either Colab or Deepnote. Keyboard shortcuts do not seem to work in an active notebook. It doesn’t seem that you can actually create notebooks in the new IDE, only run or edit existing ones.

I used Paperspace’s Gradient with its one-click notebooks almost as soon as it launched. It has been my preferred cloud notebook provider since then for a few reasons. This is because they offer many machine types (from free to $$$$) and allow you to use the native Jupyter interface. When coding locally, I use Jupyter Lab, and having a cloud option of the same means I don’t have to remember two sets of keyboard shortcuts, two UI philosophies, or two different concepts of what a notebook is.

Like the entire data science tool space, implementations of Jupyter notebooks seem to be fracturing. Perhaps this is a good development, and we will see some innovative approaches to tooling. But for me right now, none of these solutions will be replacing my preferred Jupyter Lab UI.

You can check out the new Gradient Notebook UI at https://gradient.paperspace.com/.

Thanks for reading. You can find more about me here. Consider subscribing to get notifications when I publish. If you would like to directly support my articles you can use my referral link to sign up to become a Medium member.

--

--