Get Started with Weights & Biases

Choose the right W&B product for your use case and learn how to get started

Product comparison

Welcome to Weights & Biases! Before getting started with our products, it’s important to identify which ones suit your use case.

Product	Best For	Key Features
W&B Models	Training ML models from scratch	Experiment tracking, hyperparameter optimization, model registry, visualizations
W&B Weave	Building LLM applications	Tracing, prompt management, evaluation, cost tracking for production AI apps
Serverless Inference	Using pre-trained models	Hosted open-source models, API access, model playground for testing
W&B Training	Fine-tuning models	Create and deploy LoRAs and custom model adaptations with reinforcement learning
Serverless Sandboxes	Running isolated compute environments	On-demand, disposable sandboxes for training jobs, agent tool use, and reproducible experiments

W&B Models

Models Quickstart

The “hello world” of W&B, which guides you to logging your first data.

Get Started with Models

A full-fledged tutorial that walks through the entire Models product using a real ML experiment.

W&B 101 Course

A video-led course that emphasizes experiment tracking and features quizzes to ensure comprehension.

YouTube Tutorial

Learn how models are trained, evaluated, developed, and deployed and how you can use wandb at each step of that lifecycle to help build better performing models faster.

W&B Weave

Weave Quickstart

Learn how to decorate your code so that calling into an LLM logs Weave traces and sets you on the path of a perfect LLM workflow.

Learn Weave with Serverless Inference

A full-fledged tutorial that shows Weave doing real-world evaluation of the performance of various models hosted by Serverless Inference

W&B Weave 101 Course

A video-led course that teaches you how to log, debug, and evaluate language model workflows, and features quizzes to ensure comprehension.

YouTube Demo

Learn how you can evaluate, monitor, and iterate continuously on your AI applications and improve quality, latency, cost, and safety.

Serverless Inference

Inference Introduction

Features a quickstart that shows how you use the standard OpenAI REST API to call any model hosted on Serverless Inference.

Learn Weave with Serverless Inference

A full-fledged tutorial that shows Weave doing real-world evaluation of the performance of various models hosted by Serverless Inference

Try the Inference Playground

Serverless Inference is really simple to use. Click on any model we host, start trying prompts, and see our observability layer kick into action.

Examples

Run through a few quick examples of Serverless Inference tracing calls to popular LLMs and evaluating the results.

W&B Training

Quickstart

Use W&B Training with OpenPipe’s ART library to train a model to play the game 2048.

Use your trained models

After creating your trained model, learn how to use it in your code.

Serverless Sandboxes

Sandboxes Introduction

Learn what Serverless Sandboxes are and when to use them for isolated, disposable compute.

Create a sandbox

Spin up your first sandbox in Python and start running code in seconds.

Train a PyTorch model in a sandbox

A full tutorial that walks through training a PyTorch model inside a sandbox from start to finish.

Invoke an agent in a sandbox

Give an OpenAI agent access to a sandbox for safe, isolated tool use.