Glossary · MLOps and AI infrastructure
What is Weights & Biases? Definition and meaning
A platform that logs metrics, artifacts and sweeps for machine-learning runs.
Definition
Weights & Biases: A platform that logs metrics, artifacts and sweeps for machine-learning runs.
In short: Experiment tracking and model evaluation.
When it is used
Tracking and comparing training runs.
Where it fits
Weights & Biases belongs to MLOps in the MLOps and AI infrastructure topic. These are the other terms in the same group:
- MLflow: A tool for experiment tracking, model registry and deployment.
- Kubeflow: A toolkit for pipelines, training operators and serving on Kubernetes.
- BentoML: A framework that packages models as services with dependencies and deploys them.
- KServe: A standard custom resource for serving models with autoscaling, canary rollouts and transformers.
- Ray Serve: A serving library built on Ray that composes models and scales them across a cluster.
- SageMaker: A managed service for building, training and deploying models on AWS.
- Vertex AI: A managed platform for training, tuning and deploying models and agents.
- Feature store: A system that defines, stores and serves ML features consistently for offline training and online inference.
- Model monitoring: Continuous tracking of input data, predictions and outcomes of deployed models to detect drift and performance decay.
- Model cards: Structured documentation of a model's purpose, training data, evaluation, limitations and ethical considerations.
Learn more
The full guide to Weights & Biases explains how it works, when to use it and how it compares with the alternatives, with examples. The glossary lists every term on the site from A to Z.
Previous term: Weight initialization · Next term: Whisper