Beta

Experiment Management

Track training runs with a few lines of code. Customize plots in an interactive dashboard. Automate hyperparameter optimization.

Compatible with any library, stack, or coding environment.

Make modeling a collaborative process.

The web view of your training run record. Charts show how it compares on key metrics with other runs in the same project.

Sort and filter runs to identify the most promising runs. Customize the table and save in a “View” for later.

See your results on the customizable chart dashboard. Filtering from the table of runs can be applied to charts to focus on runs of interest.

Multiple types of charts, selected to support modelers.

A training run can be instrumented to log metrics at checkpoints within a training run. These can be logged every epoch or whenever you choose.

Tracks and Runs in Model Management

Track & Organize Modeling Attributes

Modeling is messy and it can be hard to keep track of everything. With just a few lines of code, SigOpt tracks and organizes your training and tuning cycles, including: architectures, metrics, parameters, hyperparameters, code snapshots and the results of feature analysis, training runs or tuning jobs. Consider us your intern, sidekick, advisor, or all of the above.

Visualize & Compare Runs

Modeling is often about tradeoffs, but it’s hard to gather insights to properly evaluate these. Quickly gain intuition on your models and their performance with an API that automatically populates your dashboard with customizable visualizations and in-depth  hyperparameter, metric and run insights as you train and tune.

Visualization of Runs in Metric Management
Training and Tuning in Metric Management

Seamless Training & Tuning

Transitioning between training and tuning can be expensive in time and resources, so many modelers leave tuning until the last mile. Our solution fully integrates automated hyperparameter tuning with training run tracking to make this process easy and accessible. And features like automated early stopping, highly customizable search spaces, multimetric optimization and multitask optimization make tuning useful for any model you are building.

Compatible with any ML Library

Our solution is fully agnostic to modeling framework, compute stack, orchestration setup, or coding environment. And our back-end is designed to work equally well for startups with a few modelers and a few tuning jobs a month to modeling leaders with thousands of modelers and millions of concurrent tuning jobs per month.

Enterprise Platform Logo
Runs Full

Explore, Understand, Advance

We design software that sets modelers up for success. Use SigOpt to:

  • Efficiently explore your problem space by comparing results from your model development process.
  • Understand model performance with easy ways to visualize patterns in learning curves, assess parameter importance, and evaluate metric comparisons.
  • Once you’re on to something promising, advance your model with easy hyperparameter tuning before taking it into production.

Review the docs

Learn how to use Runs API through our Python client in either command line interface or notebook environments, and how to view history, visualizations, and comparisons in our dashboard.

Watch our videos

View a use case walking through how to use SigOpt for training and tuning in a fraud detection case to explore the modeling problem, understand the model options, and advance the best model to production.

Read a blog post

Get a sense for the context around why our customers requested we build this feature, how it has already made an impact for users, and how we expect it will impact your modeling workflow.

Sign Up