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.