> 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.

# How do I generate predictions on new data?

> Load a retained model and score a compatible snapshot of updated data

Load the saved model with `Trainer.load()`, reconnect to your sources, and recreate compatible table, graph, and query definitions. Call `graph.snapshot()` to materialize the updated data, then generate prediction rows using `trainer.training_table_generation_plan` and pass the snapshot to `trainer.predict()`.

The new snapshot must satisfy the trained model's target, entity, relationship, and feature requirements. Retraining is needed when you want to update the learned model, rather than score it against fresh data.

See [Recreate the Prediction Context](/kumo-engine/user-code-guide/train-a-model/publish-and-retrieve-models#recreate-the-prediction-context) and [Making Predictions](/kumo-engine/user-code-guide/making-predictions).