What the Lab is, and where your work lives

The Lab is a drawing board for physics-informed neural network (PINN) problems. You do not type code. You place blocks (a domain, an equation, some conditions, a network, and so on), connect them with wires, and the Lab turns the picture into a complete PyTorch script.

Three facts shape everything else in this tour, so it helps to have them straight from the start.

The Lab builds the script; it does not train the model

Everything on the Lab page, the canvas, the checks and the code generator, runs inside your browser tab. When you press Generate code you get a Python file. Training happens when you run that file, on your own computer or in Google Colab. That is why a problem that takes minutes to train never freezes the page, and why nothing you draw is sent to a server.

The first time you ask for a check or for code, the page loads a small Python environment into the tab. The status text reads Loading Python… and then Loading the Lab…. That is about 10 MB on the first visit and it is cached afterwards. On a very slow or metered connection the Lab waits until you actually ask for generated code, so it does not spend your data on a check you did not request.

A notebook is one problem

The Lab page is titled Your notebooks. Each notebook holds one problem: one canvas of blocks and wires, with a name. A notebook for the heat equation and a notebook for Poisson on a square are two separate notebooks.

Notebooks live in this browser

There is no account. Notebooks are saved in your browser's own storage (IndexedDB), on this device, and nowhere else. Three consequences follow.

• Clearing site data deletes them. So does a browser that refuses to store anything, such as some private windows. If the browser is blocking storage, the Lab page says so and warns that notebooks will disappear when you close the tab.
• They do not follow you to another computer by themselves. Inside a notebook, Download notebook saves a file named <notebook-name>.omega-notebook.json. On the Lab page, Open a notebook file… reads one back in and creates a new notebook from it. That is how you move a notebook to another machine, send it to a colleague, or keep a backup.
• Deleting is permanent. The Delete button on a notebook card, and Delete notebook inside a notebook, both ask you to confirm. Download first if you might want it back.

Notebooks saved by an older version of the Lab are carried forward when you open them. If the Lab had to adjust anything, it tells you how many adjustments it made and shows the first two.

What you will have by the end of the tour

1. A notebook you created yourself, from an empty canvas.
2. The wires and settings explained: what each block asks for and why.
3. A downloaded .py file, a command to run it, and a map of what is inside it.

The next page is the shortest one: creating your first notebook.