About Weights & Biases
Weights & Biases is a platform for tracking machine learning experiments and monitoring AI applications. Teams use it to compare training runs and record how their language model apps behave.
What it does
Training code sends metrics, logs and artifacts to the platform, where you can compare runs side by side. Weave, its toolkit for AI apps, records calls to language models, retrieval steps and agent steps. You add it by wrapping functions with a decorator, which keeps changes to your code small.
Key features
- Track training runs and metrics
- Compare experiments side by side
- Weave tracing for LLM calls and agent steps
- Integrations with PyTorch, Keras and scikit-learn
- Mobile app for checking training runs
Who it's for
- ML engineers tracking model training
- Teams building LLM apps that need tracing and review
- Organizations with compliance requirements
Pricing and good to know
A free plan is available, and paid tiers add team and enterprise features. The company lists ISO 27001, SOC 2 and HIPAA compliance for its platform. You can choose a multi-tenant or dedicated deployment depending on your data needs, so check which options fit your team before you sign up.
Highlights
- Track training runs and metrics
- Compare experiments side by side
- Trace LLM calls and agent steps with Weave
- Connect PyTorch, Keras and scikit-learn
- Check training runs from an iOS app
