Examples

The examples/ folder in the repository contains Jupyter notebooks demonstrating common TabuLLM workflows. They are linked here rather than rendered inline: several cells download the fraud detection dataset from Zenodo, download HuggingFace embedding models, or make paid LLM calls that require provider credentials, so they are not executed as part of the documentation build.

01_fraud_detection_walkthrough.ipynb

The core TabuLLM workflow on the fraud detection dataset:

  • TF-IDF vs. LLM embeddings

  • GMM-based dimensionality reduction with cluster quality diagnostics

  • Full ClusterExplainer usage: cost preview, outcome-based testing, per-observation diagnostics, narrative synthesis

  • A predictive pipeline combining text and structured features

External requirements: downloads the fraud dataset from Zenodo (no credentials needed); downloads a HuggingFace embedding model; uses an OpenAI model for cluster explanations (requires an OPENAI_API_KEY).

02_advanced_pipelines.ipynb

Advanced pipeline patterns:

  • Forward/backward column sweep to measure the marginal contribution of each text column

  • Stacking ensembles (single-split and multi-split) that process column groups independently and combine predictions via a meta-learner

External requirements: downloads the fraud dataset from Zenodo; downloads a HuggingFace embedding model. No LLM/provider credentials are required for this notebook.

Running the notebooks locally

pip install "tabullm[examples]"
jupyter lab examples/

See each notebook’s own setup cells for the specific credentials or model downloads it requires.