Skip to content

Track your ML experiments¶

Machine learning experimentation typically involves training models on datasets, fine-tuning hyperparameters, evaluating performance metrics, and iterating on model architectures.

Comet ML Project Page
The Comet ML Project page

With Comet, you can keep track of all your experiment iterations in a centralized platform to ensure the systematic organization, tracking, analysis, and governance of your ML experiments with built-in logging, version control, comparison, and visibility.

Comet ML Experiments¶

An Experiment is Comet's logical unit for any research workflow that you want to track and manage. Comet makes it easy for you to track your training runs by creating and running an Experiment.

Experiment Management with Comet ensures complete traceability, reproducibility, and efficient customization for your training runs, whether you're working programmatically with the Comet SDK or interactively through the Comet UI.

Experiment Management with Comet SDK¶

The Comet SDK provides the following features for experiment management:

  • Experiment object:
    • Automated logging of built-in and custom attributes within the Experiment object. Please refer to the Logging Overview for more information.
    • Support for a variety of execution flows, including online and offline execution with start, stop, and resume. Please refer to Resume an Experiment for details on the latter.
  • API object:
    • Ability to fetch any data previously logged to Comet for any experiment, project, and workspace.
    • Management functionalities to create, get, delete, and update experiments, projects, and workspaces.

Note

While the Experiment object is used for logging data during a new or resumed training, you can use the APIExperiment object for updating and getting experiment attributes in a one-off manner.

Experiment Management with Comet UI¶

The Comet UI lets you analyze and manage experiments, with all associated logged attributes, within user-friendly interfaces that you can easily share with your team members.

Discover all the functionalities of the Comet UI for experiment management from the Comet UI docs.

Example Projects with Comet ML¶

Click on the buttons below to explore example projects and get started with ML Experiment Management in the Comet UI!

Dec. 17, 2024