How to use a Jupyter Notebook on Thunder Compute

​

1. Create and connect to a Thunder Compute instance

Begin by creating and connecting to a Thunder Compute instance. Follow the instructions in our quickstart guide if you haven’t done this before.

Note: At this point, select the GPU you want to use in your notebook by running:

$ tnr device <gpu_name>

You should replace <gpu_name> with the name of the GPU you wish to utilize.

​

2. Launch the Jupyter Notebook server

Once connected to your Thunder Compute instance, start the Jupyter Notebook server by executing:

$ notebook

This command initiates a Jupyter Notebook server on the remote instance and allows you to access this server from your local computer.

​

3. Access the notebook in your browser

After running the ` command, the output will display a URL similar to:

[I 14:10:57.914 NotebookApp] Serving notebooks from local directory: /home/user
[I 14:10:57.914 NotebookApp] The Jupyter Notebook is running at:
[I 14:10:57.914 NotebookApp] http://<ip_address>:8000/tree?token=<your_token_here>
[I 14:10:57.915 NotebookApp] Use Control-C to stop this server and shut down all kernels (twice to skip confirmation).

...

    Or copy and paste one of these URLs:
        http://<ip_address>:8000/tree?token=<your_token_here>

Copy the URL provided and paste it into your web browser.

Important: Ensure the URL includes the token parameter for authentication.

​

4. Verify GPU availability in the notebook

Your Jupyter Notebook is now connected to a Thunder Compute instance with GPU capabilities. To confirm that the GPU is accessible, run the following code in a notebook cell:

import torch
print(torch.cuda.is_available())

If the setup is correct, the output should be:

true

Here is what this looks like in the browser:

That’s it! You are connected to a notebook running in a Thunder Compute instance.