DEV Community

Cover image for Jupyter Notebook v/s Google Colab
Isha Dagar
Isha Dagar

Posted on

Jupyter Notebook v/s Google Colab

I have worked on both Google Colab and Jupyter notebook so I think I would be able to explain more clearly the difference between the two and which one you should use.

Jupyter is the open source project on which Colab is based. Colab allows you to use and share Jupyter notebooks with others without having to download, install, or run anything.

Google Colab:

  • Google Colab comes with collaboration backed in the product. As Colab is a Web App hosted by Google, you need Internet connection to access it and run code.

  • It runs in cloud and runs your program on the server side so you don’t need to worry about downloading packages locally .

  • They give you a free virtual machine that has about 12 Gb of RAM which is pretty descent, but they also give access to a GPU card which is much more performing and much faster. You can access free GPUs for max 12 hours.

  • The notebooks are saved to your Google Drive account.

Jupyter Notebook:

  • Jupyter notebook run on a local host so it works offline.

  • You have direct access to your file system and with your scripts you use all the storage of your disk.

  • The hardware is your PC's RAM, Disk, CPU and GPU.

  • You have to install everything by yourself via pip or other package managers.

Jupyter notebook = Running code, taking notes , show
multimedia interactivity + Read-eval-print loop Terminal, Kernal with the frontend interfaces.

Google colab = Jupyter notebook + Collaboration, additional hardware cloud based

Which one is better?

Google colab allows you to share your work with other developers. So, not only you can push your work directly to GitHub, but you can share the notebook itself to developers and they can run the cell and get the same result as you are getting because the environment it is running on is consistent as it is a remote machine which would not be the case if you were doing a local development on jupyter notebook.
If you’re just playing around or working on personal projects, Jupyter will work fine. If you want to build commercial-grade models and deploy them to production, Codelab provides the full-lifecycle approach that you’d need.
If you are in a non-programming job, and you don't want to install everything on your work computer to get it set up for Jupyter, you can just start working with Google Colab without having to do any installation and share your scripts with non-technical co-workers who wouldn't be able to install anything themselves.

Discussion (0)