## Description Integrate W&B for comprehensive experiment tracking across all pipeline stages. ## Tasks - [ ] Data generation metrics: trajectories/hour, reward distribution, verification pass rates - [ ] Training metrics: loss curves, learning rate schedules, VRAM utilization - [ ] Evaluation metrics: benchmark scores per checkpoint - [ ] Artifact logging: best model checkpoints, trajectory samples - [ ] Hyperparameter sweep support via `wandb.sweep` - [ ] Dashboard templates for reproducibility ## Config ```python # .env WANDB_API_KEY=your-key WANDB_PROJECT=rfa-factory ```
Description
Integrate W&B for comprehensive experiment tracking across all pipeline stages.
Tasks
wandb.sweepConfig