Neptune app overview

The web app is your UI to the logged metadata. It's where you can have the data visualized, sorted, and organized.

You can watch the training live and, if needed, stop runs remotely. You can also browse tracebacks, system metrics, and source code.

How does it work?

In each project, you can organize your metadata into the following sections:

We'll walk you through each section below.

Runs

You'll typically create a run every time you execute a script that does model training, retraining, or inference. Each tracked run appears in the runs table of your project, where you can display and arrange the metadata to your liking and save views for later.

View modes

At the very top, you can toggle between three different view modes for your runs.

Runs tableRun detailsCompare runs

Choose fields (columns) that are important to you. Filter, group, or sort by any field.

Explore and preview the metadata one run at a time.

Select runs with the eye icons () to compare their metadata in this view mode.

Dashboards are fully customizable and shared project-wide

You can use customizable widgets to create dashboards in the Run details and Compare runs modes (shown in the screenshots above).

This lets you either visualize a single run in great detail or collect metadata from multiple runs into a single view.

Learn more

• Runs table
• Run details
• Compare runs
• Custom dashboards
• Custom views

Models

Models (shows all models)Model versions (shows versions of specific model)

The model registry lets you manage the metadata and lifecycle of your models separately from your experimentation runs. For each model, you can create and track model versions. To manage your model lifecycle, you can control the stage of each model version separately.

Project metadata

To facilitate collaboration, you can track metadata that applies to the whole project. For example, you could store the latest validation dataset, then access that as the source of truth for all of your runs.

Notebooks

With the neptune-notebook extension, you can snapshot and compare Jupyter Notebook checkpoints in a dedicated section of the app.

Browse examples in Neptune

The below table summarizes some points of interest, provides Neptune examples, and suggests related docs.

Component	Example	Description & docs
Runs table	View in Neptune →	The metadata of the runs organized in a table view. You can customize and save table views for later. Save custom views Group runs (groupBy)
Comparison of runs	View in Neptune →	Comparison view of selected runs. Contrast the metadata in different ways by switching between the dashboards. Compare runs: Overview
Logged metrics	View in Neptune →	Series of values are auto-displayed as charts. What you can log: Metrics and losses
Logged artifacts	View in Neptune →	Track metadata about a file or collection of files. Track artifacts Download artifact metadata
Logged system metrics	View in Neptune →	Neptune logs hardware consumption metrics by default. What you can log: System metrics
Logged interactive charts	View in Neptune →	You can display charts generated with plotting libraries in an interactive way. What you can log: Interactive visualizations
Logged images	View in Neptune →	Preview any uploaded images. What you can log: Images
Custom dashboard	View in Neptune →	You can combine different metadata types and widgets in a single view by creating custom dashboards. Create a custom dashboard
Model registry	View in Neptune →	The metadata of model versions organized in a table view. You can manage the stage of each version. Model registry overview Runs vs Model registry - Which one to use
Project-level metadata	View in Neptune →	You can also store and upload metadata common to the entire project. Log project-level metadata