How this How to Train Models track works
- Read and then write — each lesson asks for a check you can show: an input, a pass rule, and a failure case.
- Python is local — sketches run on your machine. This track does not pretend the browser trained a model.
- Prerequisites — comfort with Python and the Data Science habit of holding out data you do not train on.
- Pair with — AI for vocabulary and Generative AI when the system calls a language model.
The practical path from a frozen dataset to a checkpoint you can defend: splits, loss, updates, curves, and when to stop.
Install on your device (macOS, Linux, Windows)
Install Python 3.11+ locally for notebooks and frameworks; the on-site playground uses the dev runner when enabled.
macOS
brew install python@3.12or install from python.org (check “Add to PATH” on installers).- Create a project folder:
mkdir ~/python-practice && cd ~/python-practice. python3 -m venv .venv && source .venv/bin/activatepip install --upgrade pip
Linux
- Debian/Ubuntu:
sudo apt update && sudo apt install -y python3 python3-pip python3-venv - Fedora:
sudo dnf install -y python3 python3-pip python3 -m venv .venv && source .venv/bin/activatepip install --upgrade pip
Windows
- Install from python.org and enable Add python.exe to PATH.
- Or:
winget install Python.Python.3.12 - PowerShell:
py -3 -m venv .venv; .\.venv\Scripts\Activate.ps1 pip install --upgrade pip
Verify: python3 --version (or py --version on Windows) shows 3.11+.
Run code on this site (Backend & language playgrounds)
- Clone or open this project locally; copy
.env.exampleto.env. - Ensure
LEARNING_RUNNER_ENABLED=trueandLEARNING_RUNNER_URL=http://127.0.0.1:9999/v1/execute. - Terminal 1:
php artisan serve(orcomposer run devfor Laravel + Vite + runner together). - Terminal 2:
npm run runner— keep it running while you click Run on server.
What training actually changes on the How to Train Models track. Training changes parameters so a loss on the training rows goes down. It does not change the data, and it does not prove the model will work next month.
This is the first lesson. Read it before you change any parameters or call any tool.
The idea in practice
Write the loss, the parameter set, and the rows that are allowed to influence the update. Everything else is evaluation.
A concrete check
goal = {
'track': 'How to Train Models',
'lesson': 'What training actually changes',
}
checks = [
'input available at decision time',
'score matches the real decision',
'failure case written down',
]
print(goal['lesson'])
for item in checks:
print('-', item)
Run the sketch locally if you have Python. The printout is a reminder of the checks, not a trained model. Replace the strings with the real inputs from your own example before you treat it as a design.
What usually goes wrong
Describing training as 'showing the AI examples until it understands'. When this happens, stop adding parameters or tools. Fix the check, the data, or the permission, then run the same example again.
What to write down
- The input you are allowed to use at decision time.
- The output and the score or pass rule.
- One failure you will test on purpose.
- What you will not claim the system can do.
Practice
Name the loss and which rows are allowed to change weights.
Self-check
- Say What training actually changes in one sentence that mentions an input and an output.
- Name the failure mode in this lesson and the check that would catch it.
Done when: you can explain this lesson without the page open, and you have a written failure case.