> 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.

# Glossary

> Definitions of Kumo Tabular data, request, and prediction terms

Use this glossary for terms that have a specific meaning in Kumo Tabular and sdm-client. Follow the links for procedures, examples, and complete reference information.

## C

**Context.** The half of a `Request` where the answer is already known. The model learns from these rows. See [Make Predictions](/sdgm/kumo-tabular/user-guide/make-predictions/make-predictions).

## F

**`fit`.** Pins a request's context on the NIM (`POST /v1/sessions`) and returns a `FittedModel` for scoring several questions against the same context without re-uploading it. Takes an optional `holdout=` naming every row the session will later score. See [Make Predictions](/sdgm/kumo-tabular/user-guide/make-predictions/make-predictions#ask-the-same-context-several-questions-with-fit).

## H

**Holdout.** The `fit(holdout=...)` argument naming every row a session will go on to score, so it can be excluded from the context pinned at `fit()` time.

## I

**In-context prediction.** How Kumo Tabular answers: given labelled rows and rows to score, it predicts in one inference pass, with no training step and no per-dataset artifact.

## K

**`kumo-tabular`.** The model id sent on the wire and reported by a NIM's `/v1/models`. Named through the `KumoTabular()` handle in sdm-client.

**`KumoTabular`.** The sdm-client handle for the tabular model. Passed as the `model=` argument to `predict()`, `fit()`, and `materialize()`.

## N

**NIM.** A deployed inference service that hosts a structured-data foundation model and exposes its endpoint. sdm-client connects to a NIM and sends prediction requests; it does not deploy or run the model itself. See [Install and Connect](/sdgm/kumo-tabular/get-started/sdk-getting-started).

## P

**Predict.** The half of a `Request` naming the rows being asked about. Must not contain the target column.

**Predictive Query Language (PQL).** The language used to describe a predictive task over a `Schema`. See [Make Predictions](/sdgm/kumo-tabular/user-guide/make-predictions/make-predictions).

**`PredictionResult`.** The object every `predict()` and `fit().predict()` call returns. `to_pandas()` gives the DataFrame; upcoming explanation support exposes gradients and column cohorts through `explanations`.
See [explanation availability](/sdgm/kumo-tabular/user-guide/explain-predictions/explain-tabular-predictions). See [Prediction Results](/sdgm/kumo-tabular/user-guide/work-with-prediction-results/prediction-results).

## R

**`Request`.** What to ask, independent of which model answers it. One table with no relationships is the tabular case of the same type Kumo Relational uses. See [Make Predictions](/sdgm/kumo-tabular/user-guide/make-predictions/make-predictions).

**`row_index`.** The zero-based position of a scored row within its call, carried by every result row. Join on the entity key when present, otherwise on `row_index` — never on row order.

## S

**`SDMClient`.** The public client that owns the HTTP connection to one NIM endpoint. See [Python SDK API](/sdgm/kumo-tabular/reference/python-api-reference).

**Schema.** The table sdm-client reads, what each column means, and — for Kumo Tabular — no relationships. Built with `Schema.from_data()` from a table already in memory; `Schema.from_source()` is relational-only. See [Prepare Your Data](/sdgm/kumo-tabular/user-guide/prepare-data/data-requirements).

**Semantic type (`stype`).** Metadata describing how a column should be read, such as `ID`, `numerical`, `categorical`, `text`, or `timestamp`. See [Prepare Your Data](/sdgm/kumo-tabular/user-guide/prepare-data/data-requirements#semantic-types).

## T

**Target.** The column being predicted, named by `target=` on `Request` or by the `PREDICT` clause of a query.

**Task.** What is being asked: `classification` or `regression` for Kumo Tabular, stated explicitly rather than inferred from the target's cardinality.