Home#
Welcome to the neptune.ai docs!
Neptune is a metadata store that offers experiment tracking and model registry for machine learning researchers and engineers. With Neptune, you can log, query, manage, display, and compare all your model metadata in a single place.
-
 Try it out
See how it works with our 5-minute "Hello Neptune" example.
No registration required.
-
 Overview
What is Neptune? What can you do with it? The introduction should answer most of your questions.
-
 Set it up
Head straight to installation and setup.
Interested in Neptune on-premises?
-
 Take the tutorial
Start from the beginning, set up a sample project, and learn the central Neptune functions.
 Neptune tutorial
 Neptune R tutorial
-
What's new in the docs
Date Changes Feb 3, 2023 Added two explainer pages: using handlers for logging and using the wait()
andsync()
operations to ensure synchronous logging. Also clarified how to log from different ranks in multi-node jobs.Jan 30, 2023 Released API references that were missing for integration libraries. Published guide on using the model registry with Amazon SageMaker. Added a new page for logging charts and plots and improved discoverability of related topics. Jan 23, 2023 Added introductory guide to the Neptune-AWS integration. Added examples for logging after fitting or testing to the Lightning and fastai guides. Jan 19, 2023 Added utils
module to the API reference. Documented the newget_root_object()
method, which enhances the support for namespace handlers. Changed instances of thename
argument toproject
orworkspace
wherever names of existing projects or workspaces are passed (see neptune-client0.16.16
).Jan 17, 2023 Roughly one month ahead of the 1.0.0
release of the Neptune Python client library, we're publishing the 1.0 upgrade guide to help you understand the changes and what you'll need to update in your code.Jan 12, 2023 We're starting to roll out better Amazon SageMaker support: two new examples are now available. Jan 11, 2023 New tutorial: Working with TensorFlow. Clarified how to analyze the results in the Optuna guide. Added brief guide on how you can log datasets. Clarified how you can log 1D arrays. Jan 9, 2023 Replaced all examples of log()
withappend()
. Added a new page that collects all your options for setting your Neptune credentials in your environment. Clarified thesys/state
values. Adjusted the workaround for a Plotly/Matplotlib incompatibility issue.Looking for product release notes? See the Changelog
-
Navigating the docs
Tab Contents  Home Introduction, changelogs, installation, and setup. You'll also find information about on-premises deployment and the legacy API here.  Using Neptune Once you're all set up, in this section you'll find explanations and tips for getting the most out of Neptune. Look here for what kinds of metadata you can log and display.  Integrations Detailed guides on using Neptune together with supported libraries, tools, and environments. Neptune integrates directly with several popular ML frameworks.  Tutorials Practical walkthroughs and MLOps use cases.  API API reference material and everything you need to know about our client library. Includes Neptune CLI usage instructions.  Help If you're facing problems or have questions, we're here to help.