---
id: sagemaker
version: "3.19.0"
license: unclear
license_treatment: permissive
maintenance: active
---

# sagemaker — Open source library for training and deploying models on Amazon SageMaker.

License: permissive · Maintenance: active · Downloads: 20.9M/mo

## What it is and what it does

SageMaker Python SDK v3 is a complete rewrite of AWS's machine learning library, introducing a modular architecture split across four PyPI packages (core, train, serve, mlops) and unified APIs for training and inference. Version 3 replaces the older Estimator and Model classes with ModelTrainer and ModelBuilder, reducing boilerplate and aligning the SDK with AWS's object-oriented API design. The library supports training with PyTorch, MXNet, and Amazon's built-in algorithms, as well as custom Docker containers, and handles both distributed training and real-time inference endpoints.

The SDK is actively maintained (released 3 days ago, last commit 2026-08-14) with no known vulnerabilities and supports Python 3.10, 3.11, and 3.12. It has low install friction and is classified as Alpha (Development Status 3), indicating the v3 architecture is still stabilizing. The modular design allows you to install only the components you need (training, serving, or MLOps), though the main sagemaker package depends on all four subpackages.

Use it for:

- Train custom ML models on SageMaker using your own Docker containers or built-in frameworks like PyTorch, with data sourced from S3.
- Deploy trained models as real-time inference endpoints using ModelBuilder, replacing the V2 Model and Predictor workflow.
- Run hyperparameter tuning jobs to optimize model performance across distributed SageMaker infrastructure.
- Build end-to-end ML pipelines combining training, processing, and model registry operations via the MLOps module.
- Perform local training and inference testing before deploying to SageMaker, reducing iteration time.

## Worth the install?

AI-flagged interpretation of the facts above — verify before relying.

SageMaker Python SDK is a library for training and deploying machine learning models on Amazon SageMaker, supporting frameworks like PyTorch and MXNet as well as Amazon's built-in algorithms.

Yes, with conditions. Install if you are starting a new SageMaker project or actively migrating from v2—the modular architecture and unified APIs reduce complexity. Do not install if you have existing v2 code relying on Estimator, Model, or Predictor classes; those are unsupported in v3 and require rewriting. The package is actively maintained, permissively licensed, and has no known vulnerabilities, but it is still in Alpha (Development Status 3), so expect potential API changes in minor releases.

## Install

pip install sagemaker
uv add sagemaker
poetry add sagemaker

## Installing sagemaker

Before you install: Low install friction; the package is actively maintained with a release 3 days ago and has no known vulnerabilities. It depends on four modular SageMaker subpackages (sagemaker-core, sagemaker-train, sagemaker-serve, sagemaker-mlops) that are installed as runtime dependencies.

License in practice: Licensed under Apache Software License (permissive), which allows commercial and private use with minimal restrictions—suitable for most production deployments.

Quickstart:

pip install sagemaker

from sagemaker.train import ModelTrainer
from sagemaker.train.configs import InputData

trainer = ModelTrainer(
    training_image="my-training-image",
    role="arn:aws:iam::123456789012:role/SageMakerRole"
)
train_data = InputData(
    channel_name="training",
    data_source="s3://my-bucket/train"
)
trainer.train(input_data_config=[train_data])

Requires Python 3.10 or later; AWS credentials and IAM role configured for SageMaker access.

Verify before relying:

- Whether the modular architecture (separate PyPI packages) requires explicit installation of sagemaker-core, sagemaker-train, sagemaker-serve, sagemaker-mlops or if they are pulled automatically as dependencies.
- Compatibility and migration path for code written against SageMaker 2.x (Estimator, Model, Predictor classes are no longer supported in V3).
- Performance characteristics and resource overhead of the new unified ModelTrainer and ModelBuilder classes compared to V2 framework-specific classes.

## Package facts

- License: not declared (permissive)
- Python support: supports_current
- Install friction: low
- Maintenance: active
- Downloads: 20.9M/month (top 5,000 on PyPI)
- Known vulnerabilities: none known

## Tags

sagemaker model training deployment, aws machine learning sdk, pytorch training on sagemaker, distributed ml training aws, sagemaker inference endpoint, aws ml ops pipeline, hyperparameter tuning sagemaker, aws-integration, distributed-training, model-deployment

[View on SkillFed](https://skillfed.io/packages/sagemaker) · [View on PyPI](https://pypi.org/project/sagemaker/)