Skip to content

Project pages

This page describes more advanced procedures you can perform in Project pages in the Comet UI. Basics on Project pages are covered in Find your way around: Projects.

Create Project views

As you make changes to your current Project view, they are saved temporarily with your current URL. At any time, you can abandon the changes, or save them in a custom view, so that you can come back later and review the details in the same way.

To save a view:

  1. Click Save View.
  2. Save the view in the current view, or in a new view by selecting Create new view and giving it a name. Select Set as default view to make this the default.
  3. Click Save.

    A notification confirming the save is displayed in the window banner and the view name is displayed in the Project menu bar. The view name is research in this example:

    Project View

  4. To abandon the changes you have made, click the view name, then click Discard.

If you have made a Project visible (public), then you can share links with a particular view selected. Simply create the view and share the URL. The URL contains information about which view to use.

Info

The following items are saved with the view: selected filters, Experiment Table columns, position, and widths; Project Visualizations section expand/collapse setting and Panels. Selected areas in Panels are not saved.

To make a custom view the default for this project:

  1. Hover over the custom project view option in the list that is displayed.
  2. Click the check icon to the right of the project name to set that view as the default.

    Project View

    Now the newly saved default view will be used to determine which project view Comet renders for you. At any time, you can also delete a project view by clicking Save View and clicking Delete.

If you make changes to this view and you wish to keep them:

  1. Click Save View.
  2. Select Save changes to the view, or Create a new view.
  3. Click Save or Create.

Filter Experiments

Experiment filters allow you to select exactly which experiments to display in the Panels and Experiments views and which are shared between these two views.

Filters can be defined in either the Panels view or the Experiments view. To begin building and saving filters, open the filters sidebar using the options on the top left of the page.

Project View

You can add and save filters.

Add filters

Each filter is composed of one or multiple conditions.

To add a condition, click Add filter and select the experiment attribute you would like to set as the filter condition. Depending on the attribute’s data type, Comet exposes different operators, such as contains, is null, begins with, boolean values, and more.

Save and load filters

To save a set of filter conditions:

  1. Click Save on the top of the filters sidebar.
  2. Enter a filter name in the window that is displayed.

If you make changes to a saved filter and you wish to keep them:

  1. Click Save.
  2. Select Save changes to the filter [Filter Name], or Create new filter.
  3. Click Save.

    Project View

To load a saved filter:

  1. Open the filters sidebar.
  2. Display the saved filters in the Load Filter list.
  3. To return to a project view with all experiments, click Clear in the filters sidebar.

Perform bulk operations in the Experiment table

You can perform bulk actions on a selection of Experiments.

  1. Check the selection box at the far left of the Experiment row.
  2. Action buttons are provided along the bottom of the window. Click the action you require.

Project View

  • Archive: Soft-delete Experiments. Navigate to the Project Archive tab to either restore or permanently delete your archived Experiments.
  • Move: Move your Experiment to another Project.
  • Diff: If you select exactly two Experiments, you can select the Diff button for a detailed comparison. All aspects of the experiments are compared, including Code and Panels.
  • Tag: Add a tag to your Experiment. Select many experiments to tag them all. You can also programmatically populate tags with the Experiment.add_tag() method. To create a new tag, just enter the text and press Enter.
  • Show or Hide: Adjust your Experiment’s visibility in the Project visualizations. The Visible indicator in the Experiment Table is clickable and will either show or hide the Experiment from the Project visualizations above. You can also select the Experiments in the Project visualizations to show or hide them.
  • Stop: Lets you stop an Experiment that is running. See the next section for details.
  • Export CSV: Export your Experiment data.

In addition, you can also apply these actions to all of the Experiments in the table:

  • Group By: Group the Experiments by this column.
  • Customize Columns: Add or remove columns from the Experiment table.

Stop an Experiment

The Experiment Stop button allows you to stop one or more Experiments that are running on your computer, cluster, or on a remote system, while they are reporting to Comet. The running Experiments receive the message and raise an InterruptedExperiment exception, within a few seconds (usually less than 10).

If you don't need to handle the exception, you can simply let the script end as usual---just as if you had pressed Ctrl+C. However, if you would like to handle the interrupted script, you can do that as well. Here is an example showing a running experiment and how the exception is caught. You can perform custom code in the except clause, if you wish.

from comet_ml import Experiment
from comet_ml.exceptions import InterruptedExperiment

experiment = Experiment()

try:
    model.fit()
except InterruptedExperiment as exc:
    # handle exception here
    experiment.log_other("status", str(exc))
    # other cleanup

model.save()
experiment.log_asset("my_model.hp5")

See also the API.stop_experiment() method.

Visualize Project data with Panels

Project visualizations allow you to view and compare performance across multiple experiments.

Using Comet's Panels, you can create visualizations, tools, reports, exporters, and more.

This section describes the built-in Panels that you build entirely through the Comet UI. If those don't provide the functionality you need, you can also write custom visualizations using Python or JavaScript.

Add panels

You can add built-in Panels to enable additional functionality in your Project view.

  1. Click Add Panel in the Project visualizations area. The Panel Gallery opens: Project Visualizations
  2. Select any of the built-in, featured, or your personal collection of Panels.
  3. Make the changes you desire.
  4. Click Reset to clear all settings or Done when complete.

When you select a Panel for adding to your Project view, Comet renders a preview of the panel for you.

As described in the previous section on Experiment filters, the filters you select impact which experiments appear in the Project visualization area.

The rest of this section describes the four built-in panels. For more information on creating custom Python and JavaScript visualizations, see About custom panels.

Line charts

The Line Chart type at the project level allows you to visualize a metric (the X axis) against (the Y axis) time (duration, wall time, or step), or against another metric.

You can transform the X or Y axis using smoothing, log scale, or moving average.

At the project-level panel editor, you can set the Panel name, legend keys (experiment names or experiment keys), and outliers (show or ignore).

Project Visualizations

Bar charts

The Bar Chart type at the project level allows you to visualize a metric's aggregation (the Y axis bar height, sum by default) against experiments (the X axis).

You can change the aggregation to count, sum, average, median, mode, rms, or standard deviation.

At the project-level panel editor, you can set the Panel name, and legend keys (experiment names or experiment keys).

Project Visualizations

Scatter charts

The Scatter Chart type at the project level allows you to visualize a metric or parameter (the Y axis) against another metric or parameter (the X axis), and (through the Z axis) allow for scatter plot sizing. The value in the Z axis usually corresponds to the size of the dots in the diagram, rather than their placement.

You can transform the X, Y, or Z axis using either smoothing, log scale, moving average, minimum value, maximum value, first value logged, or last value logged.

At the project-level panel editor, you can set the Panel name, legend keys (experiment names or experiment keys), and outliers (show or ignore).

Project Visualizations

Parallel coordinates charts

The Parallel Coordinates Chart type at the project level allows you to visualize a series of metrics or parameters, where the Y axis is the metric or parameter name, and the X axis is the value of that metric or parameter. The far right vertical line is called the "Target Variable" and is typically the loss or accuracy that you are interested in. This value is often the metric that you are optimizing in a hyperparameter search.

At the project-level panel editor, you can set the Panel name.

Parallel Coordinates

Scalar Panel

The Scalar Panel allows you to aggregate your captured metrics and parameters across both Project and Experiment to truly create an interactive dashboard! The aggregated values update as new data is made available, so you can always benchmark your performance.

On the Experiment level, the aggregations are calculated on the time series you recorded for that Experiment. For example, the accuracy at each step will be used to calculate the max accuracy, or average loss.

On the Project level, the aggregations are calculated on the joined time series across all the Experiments. For example, the average accuracy is calculated by joining the accuracy values from all Experiments and averaging all of them. Values that are attributed to a specific Experiment, like min, max, first and last, are hyperlinked to that Experiment. Click on them to jump to the Experiment page!

Scalar Chart

Image Panel

The Image Panel allows you to compare your images across different experiments and throughout different steps. Through the image panel, you can search for individual images and display those images in a grid. Your rows are the individual images, while your columns are the experiment names.

Image search: search will begin after you type 3 characters. The search is for unique image names across steps and experiments. For example, if you logged image1.png in 3 different experiments across 10 different steps, your search result will still only return one result for image1.png. By selecting that image, you can view its permutations in the image panel.

Stepper: the stepper will allow you to move through the steps related to your selected images. Note that some images may be logged in some steps but not others - in those cases the panel will not display an image since none was logged.

Max number of experiments: allows you to configure how much data to use. The panel will display the images from the last X experiments from the top.

Image Panel

Data Panel

The data panel allows you to join and visualize your tabular data across your experiments. The panel easily allows you to aggregate and display your data in the panels page, sort and filter the data, and even use the filtered data output in your custom panels. The Data Panel uses unique names to aggregate the data, for example if you logged mnist_data.csv in experiment_1 and in experiment_2 you will see a single choice to select mnist_data.csv in the data panel. The panel will read both your logged csvs, join them, and visualize them as a single table.

Data selection: any data that you logged with Experiment.log_table will appear in the data selection section. It is important to use the same file name across your experiments since that is how the data is joined.

Data aggregation: the panel works in a similar way to a pandas.concat function. Much like joining two DataFrames, you will be able to specify the index of the data, the join type (inner/outer), and if you want to join the data by rows or by columns. When joining by rows, Comet will append an experiment_id column specifying where the row came from.

Data Panel

Legend values (names versus keys)

You can optionally display Experiment keys or Experiment names on the Project panel legend. For existing panels, select the vertical ellipsis in the upper corner of the panel, and select Edit Panel. When editing or creating a panel, select experiment names or experiment keys under Legend Keys.

You can edit an experiment's name by clicking on the pencil icon as you hover over the custom name column in the Experiment Table or in the single Experiment View. If you have not set a name for your Experiment, we set your Experiment Key as the default Experiment name.

For more information on setting the Experiment name programmatically, see the Experiment.set_name() method of the Python SDK.

Learn more

  • Comet Panel Gallery: A marketplace of visualizations, reports, and utilities for many of your Machine Learning needs.
  • Python Panels: Allow you to write custom visualizations the same way that you write all of your other scripts and using the Python modules that you know and love.
  • JavaScript Panels: Allow you to design, create, and use your own JavaScript methods to explore, visualize, and analyze your data.
Mar. 27, 2024