Weights Biases

About Weights Biases

Weights & Biases (W&B) is a machine learning experiment tracking and MLOps platform founded in 2017 by Lukas Biewald, Chris Van Pelt, and Shawn Lewis. W&B's core product is experiment tracking with a developer experience that won it adoption across the AI research community: a Python SDK (wandb.init(), wandb.log(), wandb.watch()) that logs metrics, hyperparameters, gradients, images, audio, video, and 3D point clouds with minimal code changes. W&B Reports provides collaborative, interactive ML analysis documents — narrative reports embedding live experiment charts used extensively in ML research papers and team retrospectives. W&B Sweeps is a distributed hyperparameter optimization system supporting grid search, random search, and Bayesian optimization with automatic early termination. W&B Artifacts provides dataset and model versioning with full lineage tracking — log input datasets, output models, and evaluation metrics in a linked DAG. W&B Launch enables running training jobs on Kubernetes, AWS SageMaker, Azure ML, and GCP Vertex AI from a unified UI. W&B Tables provides interactive dataset visualization and model evaluation — particularly strong for computer vision (image predictions, segmentation overlays) and NLP (text classification, embedding projections). W&B is the preferred experiment tracker at OpenAI, NVIDIA, Toyota, Samsung, and most leading AI research labs. Its community edition is free for individuals and academic use. W&B Model Registry was added in 2023, and Weave (LLM observability) launched in 2024 as W&B's answer to LangSmith and Arize.