PyTorchModelInference

PyTorch model inference step for prediction generation without metrics

SageMaker step type	Processing
Node type	internal (consumes upstream, produces downstream)
Container entry point	pytorch_model_inference.py
Interface file	steps/interfaces/pytorch_model_inference.step.yaml

Compute

Compute kind	framework
SDK class	PyTorch (SageMaker DLC via image_uris.retrieve)

Functionality

PyTorch model inference script. Loads trained model, preprocesses evaluation data, generates predictions without metrics computation.

Inputs (dependencies)

Input	Type	Required	Compatible producers
model_input	model_artifacts	yes	PyTorchTraining, XGBoostTraining, DummyTraining, PyTorchModel, XGBoostModel
processed_data	processing_output	yes	TabularPreprocessing, CradleDataLoading, RiskTableMapping, CurrencyConversion, BedrockProcessing

Outputs

Output	Type
eval_output	processing_output

Consumers (downstream steps)

Steps that declare this step as a compatible input source:

• ActiveSampleSelection

• ModelCalibration

• ModelMetricsComputation

• ModelWikiGenerator

• PercentileModelCalibration

• PiperMetricGeneration

• PseudoLabelMerge

• XGBoostModelInference

• XGBoostTraining

Framework requirements

Package	Version
torch	==2.1.0
transformers	==4.37.2
pandas	>=1.3.0
numpy	>=1.21.0
pydantic	==2.11.2

---

← Back to the Step Catalog