Machine Learning Experiment Tracking - Feedback

Thanks for helping us shape Machine Learning Experiment Tracking. As we explore this new feature, we would appreciate your feedback:

• General feedback
• Bugs
• Missing functionality

---

• Goal
• Getting started
  • Enabling Machine Learning Experiment Tracking on your Project
    • SaaS Users
    • Self-Managed Users
  • Usage
• Guiding principles
• Limitations

Goal

We are incubating Machine Learning Experiment Tracking, and as such we would like to hear from users feedback so that we can shape this feature into the problems they want it to solve.

Getting started

Enabling Machine Learning Experiment Tracking on your Project

SaaS Users

For SaaS Users, GitLab Model experiments are enabled for all projects.

Self-Managed Users

Requirement: Gitlab is at version 15.6 or above

Self-Managed users can enable this feature for all projects by enabling the feature flag :ml_experiment_tracking https://docs.gitlab.com/ee/administration/feature_flags.html#enable-and-disable-gitlab-features-deployed-behind-feature-flags

Usage

Usage is highlighted in the docs:

• Machine Learning Experiment Tracking
• MLFlow client integration

Guiding principles

1. Minimal setup for users, both on the Data Science side and on the platform setup side
2. Not just an additional feature, but integrated deeply into the GitLab Platform

Limitations

1. This is a project by Incubation Engineering, and we are exploring the feasibility of ML Experiment Tracking as part of the GitLab Platform. We are keeping the number of functionalities low for now as we focus on understanding use cases.
2. Further Limitations on existing functionality are described on the docs