Setup: VS Code and Google Colab

All the notebooks and labs in this course run on Google Colab, Google’s free service for running Python notebooks on its machines, some of which have a GPU. The labs are done in VS Code on your own laptop, with the notebook running on a Colab machine through Google’s Colab extension: this is how notebooks are used in research, and the labs are where you get used to it. The course notebooks that go with each session page can also be opened directly in Colab in the browser. Either way, nothing has to be installed on your laptop apart from VS Code and git, and every student works in the same environment.

What you need Why
A Google account (a free one is enough) Colab runs your code on Google’s machines
VS Code on your laptop, with the extensions Jupyter, Python and Google Colab To open, edit and run the lab notebooks
git To get all the course notebooks in one go and keep them up to date, and to keep your own work under version control (Session 2)

Every lab notebook has an Open in VS Code link (on the session page, in Canvas and in the lab quiz), and every course notebook an Open in Colab badge. All notebooks fetch their data from the course’s GitHub repository, so they run the same way everywhere. Do Parts 1 and 2 before the first lab in Session 2, and check that a notebook runs, so that the lab time is spent on the lab.

Part 1: Install VS Code, git and the extensions

  1. Install VS Code and git if you do not already have them.
  2. In VS Code, open the Extensions view (Ctrl+Shift+X, or Cmd+Shift+X on a Mac) and install Python and Jupyter (both by Microsoft) and Google Colab (by Google).
  3. Check that you can sign in to Google in a browser with the account you want to use for the course.

Part 2: Get the course notebooks into VS Code (once)

  1. Click Clone the course notebooks into VS Code. Your browser asks whether to open VS Code: allow it. (Or, in VS Code: Command Palette, Ctrl+Shift+P or Cmd+Shift+P → Git: Clone → https://github.com/ElofssonLab/kb8029-book.git.)
  2. Choose where to save the folder kb8029-book (about 100 MB), and open it when VS Code offers to. The notebooks are in its notebooks folder.
  3. Before each lab, get the latest version: Command Palette → Git: Pull. The teachers sometimes fix a lab shortly before it is used.

Part 3: Run a lab notebook on Colab from VS Code

Google’s Colab extension for VS Code (released November 2025) lets a notebook that is open in VS Code run on a Colab machine:

  1. Click the lab’s Open in VS Code link. The first time, the page that opens asks for the full path of your kb8029-book folder (for example C:\Users\you\kb8029-book or /Users/you/kb8029-book) and remembers it; after that, its button opens the notebook straight in VS Code. You can also open it yourself: in VS Code press Ctrl+P (Cmd+P on a Mac) and type the notebook’s name, for example session04-lab.ipynb.

  2. Click Select Kernel (top right) → Colab → Auto Connect (or choose a runtime with a GPU), and sign in with your Google account when asked.

  3. Run the first cells. Most notebooks start with an install cell, which installs the few packages that Colab does not already have (for example Biopython); this takes a few seconds each time you connect to a new machine.

  4. When you need a GPU (mainly Sessions 9-12), check that you got one:

    import torch
print(torch.cuda.is_available(), torch.cuda.get_device_name(0) if torch.cuda.is_available() else "")

Your changes are saved in your own copy of the notebook in kb8029-book. If Git: Pull complains that a file you changed would be overwritten, keep your version under another name (for example session04-lab-mine.ipynb) and pull again.

Things to know:

  • The code runs on Google’s machine. It cannot see files on your laptop, even though the notebook file is open in VS Code. The course notebooks download what they need from the course’s GitHub repository, so they work anyway. For your own files, upload them, or mount Google Drive (Command Palette → Colab: Mount Google Drive to Server…).
  • Colab machines are temporary. Everything on them, including installed packages and files the notebook wrote, disappears when the session ends. Your notebook itself, with its code and output, is saved on your laptop by VS Code. Save other results you want to keep (download them, or write them to Drive).
  • The free GPU is limited. A session is closed after a period of inactivity or when your quota is used up, and a GPU is not always available. Ask for a GPU only when the notebook needs one, and when you are done, release the machine with Colab: Remove Server.
  • No terminal tools. Everything in the labs is done from Python: the sequence searches and alignments in Sessions 4 and 5 call EMBL-EBI’s web services (HMMER, Clustal Omega, InterPro), and in Session 6 PyMOL runs inside the notebook and structure searches go to the Foldseek web service.

Part 4: Course notebooks in the browser

The course notebooks that go with each session page have an Open in Colab badge. Click it, choose Runtime → Change runtime type → T4 GPU if the notebook needs a GPU, and run the cells. This works from any computer. File → Save a copy in Drive keeps your own copy with your changes; without it, your edits are lost when you close the tab. You can also open a course notebook from your kb8029-book folder in VS Code, exactly as a lab.

Troubleshooting

The Open in VS Code button does nothing, or VS Code says the file does not exist. Check the folder path on the page (it must be the kb8029-book folder itself), and that the clone in Part 2 worked. Opening the notebook by name with Ctrl+P always works.

VS Code does not offer Colab under Select Kernel. Check that the Google Colab extension is installed and enabled in a local VS Code window, and reload the window (Command Palette → Developer: Reload Window).

ModuleNotFoundError. The runtime is new and the install cell has not been run in it. Run the notebook from the top.

FileNotFoundError for a file you can see on your laptop. The code runs on Colab, not on your laptop: upload the file, or read it from Drive.

No GPU, or torch.cuda.is_available() is False. You were given a machine without a GPU, or the free GPU quota is used up for now. Choose a GPU runtime explicitly, try again later, or run the notebook on the CPU: the course notebooks are small enough for that, only slower.

The connection drops in the middle of a long cell. Colab closes idle and long-running free sessions. Reconnect, run the notebook from the top, and save intermediate results to Drive if a step takes long.

Further reading