Dashboards in Neptune

A dashboard is a customizable display option for runs in your project. Dashboards are like a lens or template that applies to whichever runs are visible, which lets you quickly analyze your latest experiments and see the most relevant metadata at a glance.

You can combine different metadata types, for example:

• Overlaid metrics and loss for training and validation, on the same chart.
• Model parameters and other configuration values.
• Relationships between configs, metrics or aggregates as a scatter plot.

The dashboard displays runs that are selected with the eye icon ().

Create a dashboard

In the Runs section:

1. Navigate to Dashboards and select New Dashboard.

   If there are existing dashboards, expand the dropdown menu and then select New Dashboard.

2. Enter a title for your dashboard.

3. Click New widget and choose a widget type.

   For supported types, see Widgets and supported attribute types.

If you resize or reposition the widgets, the layout is saved as part of the dashboard configuration.

Saving a dashboard with particular runs

A dashboard configuration doesn't store which runs are visible at the time. To also save the shown runs, create a report instead.

Chart configuration

Apart from customizing each chart individually, you can configure settings to apply across all charts in the dashboard. Use Controls in the top toolbar to modify the following:

• Apply smoothing
• Change axis settings
• Change X-axis values
• Filter displayed data based on step

Note that the changes you make are saved automatically and visible to everyone who views the dashboard.

Related

Compare multiple metrics at once

Versioning

Dashboards are autosaved when you make a change.

You can access the version history through the button. Click on a version to preview the dashboard in read-only mode.

To edit share or manage a version, access the options ().

Supported widgets

The following table shows the data types that each widget supports:

Widget	Metadata type	Neptune attribute type
Chart	Numerical series (metrics, loss, accuracy)	FloatSeries
Scatter plot	Numerical or datetime values	Float, Integer, or Datetime Aggregations of metrics: Average, variance, max, min, last
Single value	Configs, scores, timestamps	Float, Integer, Boolean, String, or Datetime Aggregations of metrics: Average, variance, max, min, last

note

Displaying single values is only supported when one run is visualized on the dashboard:

• When entering the focused view by clicking on a run name.
• When only one run is selected with the eye icon ().

Exporting to report

You can export a dashboard together with particular runs as a report. This way, you can share and return to the analysis of a particular set of runs later.

1. Navigate to a dashboard and choose which runs to visualize ().

2. Select Duplicate → New report or Existing report.

   If you have an unpublished draft of an existing report, you have two options:

   • overwrite the draft with the selected widgets
   • add the selected widgets to the ones already in the report draft

3. To use the currently visible runs for the report, check the corresponding box.

4. To use the current table configuration, check the corresponding box.

   You can modify the visible runs later, while viewing or editing the report.