This tutorial covers what a lab needs from its computers and its network, what students see, and the questions they ask most. Everything in it is free, in every edition. The TA guide and the admin guide hold the details; this page is the short version for the person running the room.
The computers
- Windows lab computers use the installer. IT can install it silently on a lab image, for every user of the computer. Each student who logs in to Windows as themselves gets their own accounts and files.
- Linux labs install with uv, and students start the app with
splin a terminal. The editor opens in their browser. - A shared server can serve the app to a whole class in their browsers, with one account each. Its admin can stop a student’s run, never open it.
- Check each kind of computer before the first lab with
spl doctor. It prints one line per check, with what to do when one fails: the workspace in a synced folder, a missing WebView2, antivirus that removed a file, little memory or disk.
The network
- Nothing is sent by default. The editor talks only to the app on the same computer. There is no telemetry, no update check and no online account.
- Datasets download only when a student asks, from the Data page. The Windows installer carries MNIST, Fashion-MNIST and Palmer penguins, so Lessons 1, 2, 3, 5, 6, 7, 8 and 9 need no download at all.
- In a lab whose network is blocked, IT puts Lesson 4’s CIFAR-10 and Lesson 10’s Tiny
Shakespeare in a shared folder beforehand, with
spl datasets prefetch --to FOLDER, and setsSPELIORA_DATA_DIRS. The Data page then shows them as available.
Training on the computers you have
- Every lesson trains on a laptop’s CPU. Lesson 2 takes about 20 seconds on one core, and every other lesson less.
- One run trains at a time on a computer. A student whose run is paused for 15 minutes sees it stop by itself, and is told why.
- A GPU is used when there is one, with the Train page’s Device left on Auto. A lab that wants every run on the CPU can say so in its checks.
What students see
- The Start page, with their recent projects, the lesson they worked on last and the lessons and examples.
- The Lab page, with the lab’s checks, the locked blocks, the worksheet and Check my work.
- Problems with codes, such as
E_PORT_RANK_NEEDS_FEATURES. Each code has a help page, and many have a one-click fix, unless the lab turns fixes into explanations. - The console, a Python prompt beside the blocks, on a student’s own computer only. A lab can turn it off, as for an assessment. It never changes the blocks or the hand-in.
Check my work
Check my work is the student’s grader. It runs the lab’s checks on the model as it is on
the screen, with the same rules spl grade uses, so the marks a student sees are the marks
they get. It trains nothing, so it answers at once on any computer.
It is free in every edition, and so are the locks, the Lab page and the session summary.
Common questions
- “Nothing happens when I start it.” The window may be behind another. Otherwise the
log names the reason;
spl doctorusually finds it. - “My dataset is missing.” The Data page says so and offers Download. In a blocked lab, the shared folder is not set; ask IT.
- “I can’t change this block.” It is locked by the lab. It can be moved and renamed, never changed.
- “Check my work says 0 points, but my model trains.” Read the first failed check’s feedback: the checks mark what the lab asks for, not whether the model learns.
- “Can I start over?” Yes: start the lesson again. It makes a new project, and the old one is kept.
- “How do I hand in?” File › Export…, then Project file.