Aquanode DocsTorch + Jupyter
Deploy JupyterLab with PyTorch preinstalled as a template on a GPU machine. Aquanode provisions and starts it for you behind a private, password-protected URL.
These are starting points on a pod
Picking this in the console is how you start a pod. The app is preinstalled on the first launch, and everything you add afterwards lives on the pod's disk, so it survives pausing the machine and comes back when you resume. See Launching a pod.
The Torch + Jupyter template gives you a ready-to-use JupyterLab environment with PyTorch preinstalled, running on a GPU machine. Pick it when you deploy and it starts automatically on boot, served behind a password-protected URL. No setup on the box required.
What you can do
- Get a GPU-backed JupyterLab with PyTorch already installed
- Run notebooks for training, fine-tuning, inference, and experiments
- Access it from your browser over a private URL with an auto-generated password
Deploy Torch + Jupyter
Pick a GPU
Go to the Marketplace and choose a GPU offer sized for your workload.
Select the Torch + Jupyter template
In the deploy configuration panel, open the Template dropdown and choose Torch + Jupyter. Set a deployment name and resources, then deploy.
The service starts on boot
Aquanode provisions the box, sets up JupyterLab with PyTorch, and starts it automatically once the machine is ready. You don't install anything yourself.
Open JupyterLab
When the deployment reaches Ready, open it from Deployments and go to the Overview tab. The Jupyter section shows:
- An Open button (and the URL, with a copy button)
- A Password to sign in
Click Open, enter the password, and JupyterLab loads in your browser. PyTorch is already available, so you can import torch and start working immediately.
Access details
- JupyterLab is served on port 8888. Authentication uses a single password (there is no separate username), generated automatically per deployment and shown on the Overview tab.
- The service status (running / starting) is shown alongside the password.
The machine bills until you close it
A Jupyter deployment is a running GPU machine. It keeps billing until you close the deployment from the console. Close it when you're done.