.. _auto_aws_sagemaker_automl_predict:

*AWS SAGEMAKER AUTOML PREDICT*
==============================

Runs batch inference against AWS SageMaker AutoML-trained models by deploying endpoints (or reusing active ones), scoring the prediction dataset, and optionally tearing down endpoints and artifacts to control costs. Use this worker downstream of `aws_sagemaker_automl_learn` whenever you need to generate predictions from previously trained SageMaker AutoML models.

When to use
-----------

Tagged: ``automl``, ``aws``, ``batch``, ``cloud``, ``endpoint``, ``inference``, ``ml``, ``predict``.

Inputs
------

.. list-table::
   :header-rows: 1
   :widths: 20 20 20 20 20 20

   * - Label
     - ID
     - Type
     - Default
     - Required
     - Description
   * - Prediction Dataset
     - prediction_dataset
     - dataset
     - —
     - ✓
     - Tabular dataset to score; must have the same column structure (features) as the training dataset used in `aws_sagemaker_automl_learn` — the target column may be omitted.
   * - Models
     - models
     - dataset
     - —
     - 
     - Models metadata dataset produced by the `aws_sagemaker_automl_learn` worker, containing columns `model_name`, `target_feature`, `input_features`, and `job_name`; required when no `active_endpoints` dataset is supplied.
   * - Active Endpoints
     - active_endpoints
     - dataset
     - —
     - 
     - Previously created and still-live SageMaker endpoint metadata dataset (same schema as the `active_endpoints` output of this worker); supply instead of `models` to skip endpoint creation and reuse running endpoints.
   * - Cleanup Endpoints
     - cleanup_endpoints
     - boolean
     - True
     - 
     - Boolean flag (default: true) — when true, deletes the SageMaker endpoints after prediction completes to avoid ongoing inference costs; set to false only when you intend to reuse the endpoints via the `active_endpoints` output.
   * - Cleanup Models and Files
     - cleanup_models_and_files
     - boolean
     - False
     - 
     - Boolean flag (default: false) — when true, deletes the SageMaker model artifacts and associated S3 files after prediction; enable only if the model will not be needed for future inference runs.

Outputs
-------

.. list-table::
   :header-rows: 1
   :widths: 20 20 20 20

   * - Label
     - ID
     - Type
     - Description
   * - Predictions
     - predictions
     - dataset
     - Tabular dataset of model predictions, one row per input record, with columns identifying the target feature and its predicted value for each model.
   * - Model Prediction Information
     - model_prediction_information
     - dataset
     - Metadata dataset summarising each prediction run, including model name, job name, endpoint name, endpoint config name, input/target features, cleanup status, and any per-model errors.
   * - Active Endpoints
     - active_endpoints
     - dataset
     - Dataset describing the SageMaker endpoints that remain live after execution; only populated when `cleanup_endpoints` is false — pass this to a subsequent invocation of this worker to skip endpoint creation.
   * - Logs
     - logs
     - dataset
     - Execution log dataset with columns `log_type`, `log_message`, and `log_time` (Unix timestamp), capturing info and error events emitted during the prediction workflow.

Disciplines
-----------

- ai_ml.supervised.classification
- ai_ml.supervised.regression
- platform.integration

.. raw:: html

   <hr style="margin-top:2em">
   <p style="font-size:11px;color:#888">
   Auto-generated from <code>platform</code> schema. Worker id: <code>aws_sagemaker_automl_predict</code>. Schema hash: <code>e6ed288462c3</code>. Hand-curated docs in <code>workerexamples/</code> override this page when present.
   </p>