Course contentsShow
← Machine Learning and Deep Learning
Lesson 2816 of 3,53861. Experiment Tracking and Model ManagementFree lesson

W&B Run Management and Organization

Organizing experiments with projects, tags, groups, and notes for easy comparison and team collaboration.

W&B Run Management and Organization

What You'll Learn

You'll learn how to systematically organize your Weights & Biases experiments using projects, tags, groups, and notes to enable easy comparison and effective team collaboration.

The Challenge of Experiment Chaos

As you scale from a handful of training runs to hundreds or thousands, finding specific experiments becomes like searching for a needle in a haystack. You might wonder: "Which run used that specific learning rate?" or "What hyperparameters did my teammate use for their best model?"

Organizational Structure in W&B

Think of W&B's organizational system like a filing cabinet: projects are drawers, groups are folders within drawers, tags are colored sticky notes, and notes are detailed labels.

Projects

Projects are top-level containers that organize related work. Create separate projects for different models, datasets, or research directions (e.g., resnet-imagenet, bert-finetuning).

Groups

Groups cluster related runs within a project—perfect for hyperparameter sweeps or comparing architectural variants. All runs in a group appear together in visualizations.

Tags

Tags are flexible labels you can add or remove anytime. Use them for quick filtering: baseline, production-candidate, debugging, high-priority.

Notes

Attach rich-text notes to runs documenting insights, observations, or decisions. These become invaluable when reviewing experiments months later or onboarding new team members.

Collaboration Benefits

This structure enables powerful team workflows: your colleague can filter by experiment-v2 tag, compare all runs in the learning-rate-sweep group, and read your notes explaining why certain configurations failed—all without asking you directly.

Key Takeaway: Proper run organization with projects, tags, groups, and notes transforms W&B from a logging tool into a collaborative knowledge base that makes experiment comparison effortless and preserves institutional knowledge.