Skip to content
Products
Platform
Developer Tools
Portal

Create and manage instances of AI Model Experiments

Last updated on

Prerequisites

Section titled “Prerequisites”

Before creating an instance, ensure you meet the following prerequisites:

Understanding Instances

Section titled “Understanding Instances”

In AI Model Experiments, an instance is a dedicated MLflow™ Tracking Server environment that ensures your machine learning lifecycle is organized and secure. Each instance operates with its own independent database for metadata (parameters, metrics, tags) and dedicated storage for artifacts (models, plots, data). Therefore, instances can be used to create boundaries between different teams, projects, or stages of development.

Every instance is accessible via a unique URL. You can log in directly using your STACKIT account to view the MLflow™ UI and inspect experiments. To interact with an instance programmatically (via the MLflow™ Python SDK or REST API), you must generate access tokens as explained in Manage tokens. These tokens allow you to securely authenticate your local environments or CI/CD pipelines with the server.

To facilitate artifact storage, a dedicated object storage bucket is automatically generated in your project for every instance you create. Because instances rely on these buckets, you must have Object Storage enabled in your project before you can enable AI Model Experiments.

To keep your storage costs optimized and your workspace clean, instances include a custom setting called DeletedExperimentRetention. When you delete an experiment in the MLflow™ UI or via the API, the data is initially only “marked for deletion” (Soft Delete). This setting defines the grace period before that data is permanently deleted from the database and artifact storage. If you do not specify a value, the retention period defaults to 30 days. You can adjust this timeframe to meet your team’s specific data-handling policies or to reclaim disk space.

Create an Instance

Section titled “Create an Instance”
  1. On the sidebar click on AI Model Experiments.
  2. On the top bar click on Create AI Model Experiments.
  3. Enter an Instance name and optionally a Description and a DeletedExperimentRetention.
  4. Click on Order fee-based.

List Instances

Section titled “List Instances”
  1. On the sidebar click on AI Model Experiments.
  2. You can see a list of all instances. Click on an instance to view the details pane.

Update Instances

Section titled “Update Instances”
  1. On the sidebar click on AI Model Experiments.
  2. You can see a list of all instances. Click on an instance to view the details pane.
  3. Modify the desired entry.
  4. Click on Save to update the instance.

Delete Instances

Section titled “Delete Instances”
  1. On the sidebar click on AI Model Experiments.
  2. You can see a list of all instances. Click on an instance’s three dots button and then on Delete to open the delete pane.
  3. Enter the instance’s name and click on Delete to finally delete the instance.