Version: v0.23.x

Experiments

Overview

The Experiments page shows all fine-tuning jobs in your current project. From here you can view experiment status, monitor training metrics, and create new experiments from scratch or based on previous runs.

You can open the Experiments page from:

The Experiments card on the project homepage
The Experiments tab in the top navigation bar

Project homepage with the Experiments card and the Experiments tab highlighted

This section covers:

Create a new experiment
View and manage an experiment
Advanced configuration
Problem-type details: Text classification, Image classification, and Object detection

Experiment list view

Each row in the experiment list shows:

Name
Experiment ID (hidden by default)
Dataset
Created date
Status: queued, starting, training, validating, completed, failed, or cancelled

Use the column toggle at the top right to show or hide columns, including the Experiment ID and validation metric columns.

You can:

Search using the search bar under the Name, Experiment ID, and Dataset column headings
Sort by any column (newest first by default)
Click the select icon (☑) at the top left of the table to select multiple experiments, then Delete or Compare them, or use Ask KGM if at least one selected experiment is completed
Click ⋮ on a row to open its action menu: Copy, Rerun, and Delete, plus Deploy, Predict, Download Validation Results, Ask KGM, and Create AutoML for completed experiments, or Stop while an experiment is running