> This page is for Structured Data Models Client.

> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.nvidia.com/sdgm/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.nvidia.com/sdgm/_mcp/server.

# Notebooks

> Follow an example Kumo Tabular workflow in a notebook

Download this quickstart notebook to learn how Kumo Tabular uses labeled context
rows to predict outcomes for new rows without training a task-specific model.
The example classifies breast masses as benign or malignant using four features
from the Mammographic Mass dataset on Kaggle.

In the notebook, you will:

* Install and configure `structured-data-models-client`.
* Connect to a deployed Kumo Tabular NVIDIA NIM inference service.
* Read the sample CSV from a Snowflake stage and prepare labeled context and test rows.
* Build and inspect a schema, then submit a classification request.
* Read class probabilities and compare accuracy and ROC AUC with a majority-class baseline.
* Run the same task using Predictive Query Language (PQL).

Before you begin, you need a deployed Kumo Tabular NIM endpoint and an active
Snowpark session. Download the
[Mammographic Mass dataset from Kaggle](https://www.kaggle.com/datasets/overratedgman/mammographic-mass-data-set),
get the `Cleaned_data.csv` file, and upload it to a Snowflake stage. Use a notebook
environment with pandas, NumPy, scikit-learn, and requests available. The model
runs on the NIM server, so the notebook machine does not need a GPU.

For client installation and connection requirements, see
[Install and Connect](/sdgm/kumo-tabular/get-started/sdk-getting-started).

Download the Kumo Tabular quickstart notebook