Weights & Biases

Weights and Biases is a platform for experiment tracking, dataset versioning, model registry, and visualization across machine learning pipelines. The wandb Python library logs metrics, hyperparameters, gradients, and artifacts from training runs to a hosted or self-hosted server with a web UI for comparisons and reports. The team also ships Weave, a companion library focused on tracing, evaluation, and monitoring of LLM and agent apps.