Weights & Biases API

Programmatic logging of training metrics, hyperparameters, and metadata to a centralized cloud or self-hosted backend via the Python SDK or REST API. Metrics are persisted with timestamps and run context, enabling real-time visualization dashboards and historical comparison across experiments. The system automatically captures framework-specific integrations (PyTorch, TensorFlow, scikit-learn) to reduce boilerplate logging code.

Automated hyperparameter search via Bayesian optimization, grid search, or random search configured through a YAML sweep specification. The system launches parallel training jobs across local or cloud compute, logs metrics for each trial, and recommends optimal hyperparameters based on a user-defined objective (e.g., maximize validation accuracy). Supports conditional parameters, nested search spaces, and early stopping to reduce wasted compute.

W&B provides a query expression language (documented in 'Query Expression Language' section) enabling programmatic filtering and aggregation of experiment runs, metrics, and artifacts. Queries are executed via Python SDK or REST API, returning structured results for analysis, reporting, or automation. Supports complex filters (e.g., 'accuracy > 0.9 AND learning_rate < 0.01') and aggregations (e.g., 'max accuracy per hyperparameter').

W&B SDK provides framework-agnostic integration with popular ML libraries (PyTorch, TensorFlow, scikit-learn, XGBoost, Hugging Face Transformers, etc.) via auto-logging that intercepts native logging calls and framework hooks. Users add minimal boilerplate (e.g., `wandb.init()`, `wandb.log()`) to enable automatic metric capture, model checkpointing, and hyperparameter logging without modifying training code. Supports custom integrations via decorators and callbacks.

W&B supports team-based access control with role-based permissions (admin, member, viewer) and project-level sharing. Teams can be created in cloud tier (Pro and above) or self-hosted Enterprise tier. Access control enables fine-grained sharing of experiments, models, and reports with team members or external stakeholders. Audit logs (Enterprise tier) track all data access and modifications for compliance.

W&B Personal tier (free) and Enterprise tier support self-hosted deployment via Docker, enabling on-premise installation for teams with data residency or security requirements. Self-hosted instances run independently from W&B cloud, with optional integration to W&B cloud for cross-instance features. Supports custom domain configuration, HTTPS, and integration with corporate identity providers (LDAP, SAML, OAuth).

Centralized model artifact storage with versioning, lineage tracking, and metadata tagging. Models are stored as W&B Artifacts (immutable, content-addressed files) linked to specific experiment runs, enabling reproducibility by pinning a model version to its training config and metrics. Supports model comparison, promotion workflows (dev → staging → production), and integration with CI/CD pipelines for automated model deployment.

Framework for evaluating LLM outputs against custom scoring functions and datasets. Users define evaluation logic (e.g., BLEU score, semantic similarity, custom classifiers) that runs on model predictions, generating structured evaluation reports. Integrates with W&B Weave for tracing LLM calls and with W&B Models for comparing evaluation results across model versions. Supports batch evaluation of large datasets and cost estimation for LLM API calls.

W&B Weave provides distributed tracing for LLM applications, capturing function calls, LLM API requests, and intermediate outputs in a queryable trace tree. Traces are visualized as DAGs showing data flow through the application, enabling debugging of multi-step LLM pipelines (e.g., RAG systems, agents). Integrates with OpenAI, Anthropic, and other LLM providers to auto-capture API calls without code changes. Supports cost tracking and latency profiling per trace.

W&B Artifacts system enables versioning of datasets as immutable, content-addressed files linked to experiments. Datasets are tagged with metadata (e.g., 'train-v2.3', 'test-split-1') and tracked through the ML pipeline, creating a lineage graph showing which models were trained on which dataset versions. Supports dataset comparison (schema changes, row count diffs) and integration with data processing workflows to track transformations.

W&B integrates with CI/CD systems (GitHub Actions, GitLab CI, Jenkins) to trigger model training, evaluation, and deployment workflows based on code or data changes. Supports conditional execution (e.g., 'only run sweep if accuracy improved'), automated alerts (Slack, email) on metric thresholds, and promotion workflows that move models through dev → staging → production with approval gates. Webhook system enables custom automation logic.

W&B Reports enable creation of interactive dashboards combining experiment metrics, model comparisons, and custom visualizations (plots, tables, markdown). Reports are shareable via web links with fine-grained access control (view-only, edit, admin). Supports embedding reports in documentation, exporting to PDF, and version history for collaborative editing. Reports automatically update when underlying experiment data changes.

W&B Inference provides an OpenAI-compatible API for accessing open-source foundation models (Llama, Mistral, etc.) without managing infrastructure. API supports streaming responses, token counting, and usage tracking integrated with W&B cost monitoring. Requests are routed through W&B's hosted infrastructure or can be self-hosted. Supports both chat completions and text completions endpoints compatible with OpenAI SDK.

W&B Training (preview) enables serverless fine-tuning and post-training of open-source LLMs using reinforcement learning and supervised fine-tuning. Users provide training data and configuration; W&B handles compute provisioning, distributed training, and checkpointing. Supports multi-turn agentic task training for building task-specific models. Results are automatically versioned and integrated with W&B model registry.