Skip to content

Model kinds notebook

A marimo notebook that fits a binary classifier, a multiclass classifier and a regressor on the same features of the iris data. It shows the kind that model_kind reads for each model, how the scores become the output, the graph text of each kind, the local explanations of each kind, and the error for a target class that the model cannot reach.

  • Everything up to "Put it in words" runs without an API key and sends no request. The iris data comes with scikit-learn, so the notebook downloads nothing.
  • The last cells describe the petal length graph of each model with the LLM, and only run when you press their button. The default model needs OPENROUTER_API_KEY. Pass model= to use another model, see Choosing the model.

The page below is the notebook with the output of its cells, exported when the docs are built. The buttons wait for a press, so they show their idle state here. To run them:

uv run marimo edit notebooks/model_kinds.py

The usage guide explains the graph options, and the local explanations guide explains the result types.

Open the notebook in its own page