Weights & Biases

Weights & Biases (W&B) is the leading platform for ML experiment tracking, providing tools to log, visualize, and compare model training runs across your team. It automatically captures everything needed to reproduce any experiment.

Key use cases include:

• Experiment tracking — log hyperparameters, metrics, loss curves, and system stats automatically
• Run comparison — visually compare hundreds of experiments side by side
• Model registry — version and stage models through development to production
• Artifact management — track datasets, model checkpoints, and evaluation results
• Sweeps — automated hyperparameter optimization with Bayesian search
• LLM fine-tuning tracking — log prompts, completions, and eval metrics for LLM workflows

W&B is used by ML teams at companies of all sizes (from solo researchers to OpenAI, NVIDIA, and Microsoft) who need to move beyond spreadsheets and ad-hoc experiment management. It integrates seamlessly with PyTorch, Hugging Face Transformers, Keras, and every major training framework.

1. Create a free account at wandb.ai.
2. Install and authenticate:

   pip install wandb
   wandb login
   # Paste your API key from wandb.ai/authorize

3. Add tracking to your training script:

   import wandb
   wandb.init(project='my-project')
   for epoch in range(10):
       loss = train_one_epoch()
       wandb.log({'loss': loss, 'epoch': epoch})
   wandb.finish()

4. View results in the dashboard at wandb.ai.

Pricing: Free for personal use (unlimited experiments). Teams plan starts at $50/user/month with collaboration features. Enterprise pricing is custom. See wandb.ai/pricing.