sagemaker-experiments

What it is and what it does

sagemaker-experiments is a Python SDK for tracking machine learning experiments within AWS SageMaker. It provides a high-level interface to organize and log experiment metadata—experiments (collections of trials), trials (multi-step workflows), and trial components (individual steps like data cleaning or model training)—directly from Python scripts, notebooks, and SageMaker jobs. The package depends on boto3 and offers a Tracker context manager for logging trial component information.

The package is designed to integrate with SageMaker Studio and enable comparison of multiple training runs. However, the description explicitly notes that the official SageMaker SDK is now the recommended path and this repository may not reflect the latest product improvements. It remains useful for legacy workflows or direct experiment tracking in existing SageMaker setups.

Use it for:

• Log and organize multiple training runs in SageMaker notebooks to compare hyperparameters and model performance.
• Track multi-step ML workflows where each step (preprocessing, training, evaluation) is a separate trial component.
• Record experiment metadata from inside running SageMaker training and processing jobs for later analysis.
• Build experiment lineage and audit trails for reproducibility in production ML pipelines.
• Query and analyze experiment results using the ExperimentAnalytics interface to extract metrics across trials.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Tracks machine learning experiments, trials, and trial components in AWS SageMaker training jobs, processing jobs, and notebooks.

Yes, if you are already using SageMaker and need lightweight experiment tracking in existing jobs or notebooks. No, if you are starting a new SageMaker project—use the official SageMaker SDK instead, as this package is dormant (last update 2023-05-17) and the description recommends the SDK for up-to-date features. Install only for legacy codebases or when you specifically need the standalone Tracker interface.

Install

sagemaker-experiments on PyPI

pip install sagemaker-experiments

uv add sagemaker-experiments

poetry add sagemaker-experiments

Installing sagemaker-experiments

Before you install

Low install friction with a single runtime dependency (boto3). Maintenance is dormant—last commit was 2023-11-14, over 1185 days ago—but the package is marked Production/Stable and the description explicitly recommends using the SageMaker SDK instead for up-to-date features.

License in practice

Licensed under Apache License 2.0 (permissive), so you can use, modify, and distribute it freely in commercial and private projects without restriction.

Quickstart

pip install sagemaker-experiments

import boto3

# Create and manage experiments via the package's API
# See documentation for Experiment.create() and trial management

Requires AWS account credentials configured for boto3 and IAM permissions for SageMaker.

Verify before relying

• Whether this package receives security patches or bug fixes despite dormant status since 2023-11-14.
• Current compatibility with recent SageMaker SDK versions and whether the official SDK is now the recommended path for all use cases.
• Specific API methods and usage patterns beyond what the description excerpt provides.

Package facts

Evidence: sagemaker_experiments-0.1.45-py3-none-any.whl

Keywords: ML, Amazon, AWS, AI, Tensorflow, MXNet

Tags

More Artificial Intelligence packages