Skip to main content

Documentation Index

Fetch the complete documentation index at: https://wb-21fd5541-john-wbdocs-2044-rename-serverless-products.mintlify.app/llms.txt

Use this file to discover all available pages before exploring further.

Browse end-to-end examples grouped by product. Each page below links to a tutorial, walkthrough, or page containing a runnable colab notebook.

Quickstarts

Get started with W&B

Cross-product entry point for getting up and running with Weights & Biases.

W&B Models: Quickstart

Install W&B and start tracking experiments in minutes.

W&B Models: Get started with experiments

Track experiments, log metrics, and visualize results.

W&B Weave: Quickstart

Add tracing to your LLM application to debug and monitor model interactions.

W&B Weave: Learn Weave with Serverless Inference

Trace model calls, compare outputs, and run evaluations using Serverless Inference.

W&B Weave: Leaderboard quickstart

Build a leaderboard to compare models and experiments.

W&B Models

Experiments

W&B Models: Experiments overview

Intro to tracking metrics, hyperparameters, system metrics, and model artifacts.

W&B Models: Configure experiments

Save experiment configuration with a dictionary-like object.

W&B Models: View experiment results

Explore run data in interactive workspaces.

W&B Models: Log media and objects

Log rich media: 3D point clouds, molecules, HTML, histograms, and more.

W&B Models: Log models

Log model artifacts to a run with run.log_model() and run.use_model().

W&B Models: Create and track plots

Build and track plots from ML experiments.

Sweeps, artifacts, tables

W&B Models: Sweeps overview

Intro to hyperparameter search and model optimization with Sweeps.

W&B Models: Run a sweep

Define, initialize, and run a hyperparameter sweep.

W&B Models: Artifacts overview

Intro to W&B Artifacts and how to get started.

W&B Models: Create and use a dataset artifact

Create, track, and use a dataset artifact across experiments.

W&B Models: Manage artifact retention

Set TTL policies on artifacts to manage storage.

W&B Models: Tables overview

Iterate on datasets and understand model predictions with Tables.

W&B Models: Log tables and query data

Log tables, visualize, and query structured data.

W&B Models: Visualize and analyze tables

Compare, filter, group, and sort tables in merged or side-by-side views.

Registry, reports, UI features

W&B Models: Registry overview

Manage and share artifact versions across your organization.

W&B Models: Reference an artifact version with aliases

Use default, custom, and protected aliases in the Registry.

W&B Models: Reports overview

Project management and collaboration tools for ML projects.

W&B Models: Create a report

Create a W&B Report with the App UI or programmatically.

W&B Models: Edit a report

Edit reports interactively or with the Report API.

W&B Models: Custom charts overview

Build custom charts in W&B projects with Vega visualizations.

W&B Models: Build custom charts

Use custom charts to build tailored visualizations in the W&B UI.

W&B Models: Embed objects

Use the embedding projector to explore object embeddings.

ML framework integrations

W&B Models: Keras

Track experiments, checkpoint models, and visualize predictions with Keras callbacks.

W&B Models: PyTorch

Track metrics, gradients, and models with the PyTorch integration.

W&B Models: PyTorch Lightning

Use the built-in WandbLogger with PyTorch Lightning.

W&B Models: PyTorch Ignite

Automatically log training metrics, model parameters, and configs.

W&B Models: PyTorch torchtune

Track LLM fine-tuning experiments with the torchtune WandBLogger.

W&B Models: TensorFlow

Log custom metrics, use estimator hooks, and sync TensorBoard logs.

W&B Models: XGBoost

Log gradient boosting metrics, feature importance, and model performance.

W&B Models: YOLOv5

Use the built-in W&B integration in YOLOv5 for experiment tracking and versioning.

ML library integrations

W&B Models: Hugging Face

Visualize and track Hugging Face model performance with W&B.

W&B Models: Hugging Face Transformers

Use W&B with the Hugging Face Transformers Trainer.

W&B Models: Simple Transformers

Integrate W&B with Hugging Face Simple Transformers.

W&B Models: Hugging Face Diffusers

Autolog prompts, generated media, and pipeline architecture.

W&B Models: OpenAI API

Log chat completions, fine-tuning jobs, and token usage metrics.

W&B Models: Azure OpenAI fine-tuning

Fine-tune Azure OpenAI models with W&B experiment tracking.

W&B Weave

Tutorials

W&B Weave: Build an evaluation

Build an evaluation pipeline with Weave Models and Evaluations.

W&B Weave: Evaluate a RAG application

Build and evaluate a RAG application with LLM judges.

W&B Weave: Trace nested functions

Track deeply nested call structures with W&B tracing.

W&B Weave: Version an application

Track and version your application and its parameters with Weave Model.

Cookbooks

W&B Weave: Introduction to traces

A beginner-friendly introduction to tracing with Weave.

W&B Weave: Introduction to evaluations

Get hands-on with running evaluations in Weave.

W&B Weave: Hugging Face dataset evaluations

Run evaluations on Hugging Face datasets with Weave.

W&B Weave: Import a dataset from CSV

Load a CSV into a Weave dataset and use it in evaluations.

W&B Weave: Use Weave with W&B Models

Combine W&B Models and Weave in a single workflow.

W&B Weave: Chain of density summarization

Implement chain-of-density prompting for iterative summarization.

W&B Weave: DSPy prompt optimization

Optimize prompts with DSPy and track results in Weave.

W&B Weave: NotDiamond custom routing

Route between models dynamically with NotDiamond.

W&B Weave: Multi-agent structured output

Coordinate multiple agents that produce structured output.

W&B Weave: Code generation

Build and evaluate a code-generation pipeline with Weave.

W&B Weave: OCR pipeline

Trace and evaluate a computer-vision OCR pipeline.

W&B Weave: Audio with Weave

Work with audio inputs and outputs in Weave traces.

W&B Weave: Online monitoring

Monitor a production LLM application with Weave.

W&B Weave: Production feedback

Collect and act on user feedback from production traffic.

W&B Weave: Scorers as guardrails

Use Weave scorers as guardrails for production LLM calls.

W&B Weave: Custom model costs

Track custom per-model costs alongside traces.

W&B Weave: PII data handling

Redact PII in Weave traces for sensitive workloads.

W&B Weave: Use the Weave Service API

Call the Weave Service API directly to record traces.

Evaluation and tracking

W&B Weave: Evaluate using local scorers

Use small local language models to evaluate AI safety and quality.

W&B Weave: Set up annotation queues

Route traces to domain experts and export structured feedback.

W&B Weave: Track custom costs

Track and manage costs for LLM operations.

LLM provider integrations

W&B Weave: Anthropic

Automatically track and log LLM calls made via the Anthropic SDK.

W&B Weave: Cohere

Automatically track and log LLM calls made via the Cohere Python library.

W&B Weave: Google

Trace and log Google GenAI model calls.

W&B Weave: Groq

Track and monitor Groq LPU inference with Weave.

W&B Weave: MistralAI

Trace and evaluate Mistral AI model calls with Weave.

W&B Weave: OpenAI

Integrate OpenAI with Weave for tracing, evaluation, and monitoring.

W&B Weave: LiteLLM

Automatically track and log LLM calls made via LiteLLM.

Framework and protocol integrations

W&B Weave: CrewAI

Monitor and trace multi-agent applications with CrewAI.

W&B Weave: DSPy

Track and log calls made using DSPy modules and functions.

W&B Weave: Instructor

Trace structured-output calls made via Instructor.

W&B Weave: LangChain

Track and log all calls made through the LangChain Python library.

W&B Weave: Verdict

Use the Verdict evaluation framework to monitor LLM evaluation pipelines.

W&B Weave: Hugging Face Hub

Track and analyze ML applications with Hugging Face Hub.

W&B Weave: Model Context Protocol (MCP)

Trace activity between your MCP client and MCP server.

Serverless Inference

Serverless Inference: Create a fine-tuned LoRA

Fine-tune and deploy a LoRA adapter with Serverless Inference.

Serverless Inference: Use Cline with Serverless Inference

Integrate Cline with the Serverless Inference endpoints.

W&B Training

W&B Training: Serverless RL

Post-train models with reinforcement learning on W&B.

W&B Training: Use Serverless SFT

Fine-tune models with Serverless SFT using the OpenPipe ART framework.

W&B Training: Use trained models

Make inference requests to the models you’ve trained.

Serverless Sandboxes

Serverless Sandboxes: Train a PyTorch model

Train a PyTorch model in a Serverless Sandbox environment.

Serverless Sandboxes: Invoke an agent in a sandbox

Invoke an OpenAI agent within a Serverless Sandbox.