XGBoostTraining

XGBoost model training step

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

Compute

Compute kind	estimator
SDK class	XGBoost (SageMaker DLC via image_uris.retrieve)

Functionality

XGBoost training for tabular classification with risk table mapping and numerical imputation. Supports binary/multiclass, class weights, pre-computed artifacts, and comprehensive evaluation metrics.

Inputs (dependencies)

Input	Type	Required	Compatible producers
input_path	training_data	yes	TabularPreprocessing, BedrockProcessing, StratifiedSampling, RiskTableMapping, MissingValueImputation, LabelRulesetExecution, ProcessingStep, DataLoad, PyTorchModelInference
hyperparameters_s3_uri	hyperparameters	no	HyperparameterPrep, ProcessingStep
model_artifacts_input	processing_output	no	XGBoostTraining, MissingValueImputation, RiskTableMapping, FeatureSelection

Outputs

Output	Type
model_output	model_artifacts
evaluation_output	processing_output

Consumers (downstream steps)

Steps that declare this step as a compatible input source:

• DataUploading

• DummyTraining

• LightGBMMTModelInference

• LightGBMModelEval

• LightGBMModelInference

• ModelCalibration

• Package

• Payload

• PercentileModelCalibration

• PyTorchModelEval

• PyTorchModelInference

• TSAModelCalibration

• XGBoostModel

• XGBoostModelEval

• XGBoostModelInference

• XGBoostTraining

• XgboostMtModelEval

Framework requirements

Package	Version
xgboost	==1.7.6
scikit-learn	>=0.23.2,<1.0.0
pandas	>=1.2.0,<2.0.0
pyarrow	>=4.0.0,<6.0.0
boto3	>=1.26.0
pydantic	>=2.0.0,<3.0.0
matplotlib	>=3.0.0
numpy	>=1.19.0

---

← Back to the Step Catalog