Cascade UI#

Cascade UI is a lightweight dashboard for your machine learning experiments. The tool provides a visual overview of your experiment results, metadata, parameters, metrics and more.

This single interface is a replacement for dash-based Viewers, that were used previously to visualize results.

Installation#

Install the latest version using pip

pip install cascade-ui

Note that UI support in Cascade starts with the version 0.16.0.

Basic Usage#

You can run Cascade UI with a simple CLI command. Navigate to the Workspace folder or a parent folder to your repo and run.

cascade ui

The command will open a server on local port 8000. If you open a link in your browser you will see your workspace overview.

Workspace page of Cascade - Small scale MLOps library

You can interact with the workspace itself or go deeper into any repo.

Repos#

Repo page features a table of lines with their basic info.

Repo page of Cascade - Small scale MLOps library

Lines#

Line page features customizable table with the info about models, comments and plots.

Line page of Cascade - Small scale MLOps library

Select columns#

You can select columns from the list and request them from model’s meta.

Table customization page of Cascade - Small scale MLOps library

Plots#

Inside each line you can visualize the change of metrics.

Models#

Model page provides detailed overview of machine learning experiment metadata, tracked parameters and metrics.

Model page of Cascade - Small scale MLOps library

Configs#

Here you can see configs produced by Cascade’s configuration management system. If your model has a config saved it will be displayed here.

Model config page of Cascade - Small scale MLOps library

Logs#

If you used cascade run with log tracking, you will be able to see your logs here.

Comments#

You can comment on each container using UI.