Beginner

Introduction to Weights & Biases

Discover the ML developer platform used by OpenAI, NVIDIA, and thousands of teams for experiment tracking, model optimization, and collaboration.

What is Weights & Biases?

Weights & Biases (W&B) is a machine learning platform that helps ML practitioners track experiments, visualize results, optimize hyperparameters, and share findings. Founded in 2017, it has become one of the most popular tools in the ML ecosystem.

W&B is a SaaS platform with a generous free tier for individuals and academic teams. Enterprise customers can deploy it on-premises or in a private cloud.

Core Products

📊

Experiments

Track metrics, parameters, code, system metrics, and media (images, audio, video) with just a few lines of code.

🔧

Sweeps

Automated hyperparameter optimization with Bayesian, grid, and random search. Distributed across multiple machines.

📦

Artifacts

Version and track datasets, models, and other files. Build lineage graphs to understand data flow.

📚

Reports

Create interactive documents combining live charts, markdown, and code. Share findings with your team.

W&B vs Alternatives

FeatureW&BMLflowNeptuneClearML
HostingSaaS + EnterpriseSelf-hosted (free)SaaS + EnterpriseSaaS + Self-hosted
TrackingExcellentExcellentExcellentGood
VisualizationBest-in-classGoodGoodGood
SweepsBuilt-in (Bayesian)Via integrationsVia integrationsBuilt-in
CollaborationExcellent (Reports)LimitedGoodGood
Free tierGenerous (unlimited runs)Fully freeLimitedGenerous
Model ServingVia RegistryBuilt-inNoBuilt-in
✅
When to choose W&B: W&B is ideal when visualization quality, collaboration features, and ease of use are priorities. It excels for deep learning research teams. Choose MLflow if you need self-hosted and open-source. Choose W&B if you want the best UI and don't mind SaaS.

How It Works

The W&B workflow is simple:

  1. Add a few lines of code

    Import wandb, initialize a run, and log metrics. That's it — 3 lines to get started.

  2. View results in the dashboard

    Real-time charts update as your training progresses. Compare runs side-by-side.

  3. Optimize with Sweeps

    Define a search space and let W&B find the best hyperparameters automatically.

  4. Share via Reports

    Create interactive reports with live visualizations and share with your team.

Python — W&B in 5 lines
import wandb

wandb.init(project="my-first-project")
wandb.config.learning_rate = 0.001

for epoch in range(100):
    loss = train_one_epoch()
    wandb.log({"loss": loss, "epoch": epoch})

wandb.finish()
💡
Prerequisites: Basic Python knowledge and familiarity with machine learning concepts. A free W&B account (sign up at wandb.ai). No credit card required for the free tier.