wandb/wandb

What is it?

Key features

• 01Log hyperparameters and metrics from a Python script with wandb.init() and run.log().
• 02Compare and visualize runs in a web dashboard.
• 03Integrates with PyTorch, TensorFlow, Keras and JAX.
• 04Version datasets and models as artifacts.
• 05Run hyperparameter sweeps to search configurations.
• 06Deploy as cloud, dedicated cloud or self-managed server.

Use cases

• Tracking training runs and comparing model variants during research.
• Versioning datasets and model checkpoints across a team's ML pipeline.
• Running hyperparameter sweeps for deep learning models.
• Hosting an experiment tracking server inside a private network.

Quick start

1. 1

   Install the wandb library

   pip install wandb
2. 2

   Sign up and create an API key at wandb.ai/settings (or run wandb login)

   wandb login
3. 3

   In your Python script, initialize a run with wandb.init() and log metrics with run.log()

Things to know

• !Full features rely on a W&B account and API key, with the hosted platform as the default.
• !Self-hosting W&B Server is a separate deployment effort beyond the Python library.

Topics

FAQ

What is wandb?

Python library and hosted platform for tracking, visualizing and managing machine learning experiments, datasets and models. wandb is the Python client for Weights & Biases, a platform for recording machine learning work. Inside a training script or notebook, you start a run with wandb.init(), pass in hyperparameters as a config dictionary, and log metrics such as accuracy and loss with run.log(). The results appear in a web dashboard where you can compare runs and see how values changed step by step.

Is wandb free to use? What license does it have?

wandb is open source under the MIT license. Check the license terms before using it in commercial products.

Is wandb still maintained?

Actively maintained. The last commit to wandb was on Oct 22, 2026.

What can I use wandb for?

Tracking training runs and comparing model variants during research. Versioning datasets and model checkpoints across a team's ML pipeline. Running hyperparameter sweeps for deep learning models. Hosting an experiment tracking server inside a private network.

How do I install wandb?

Install the wandb library: pip install wandb Sign up and create an API key at wandb.ai/settings (or run wandb login): wandb login In your Python script, initialize a run with wandb.init() and log metrics with run.log()

Where can I get wandb?

The source code and installation instructions are on GitHub: https://github.com/wandb/wandb