Track robotics training dynamics in Weights & Biases

Weights & Biases helps accelerate robotics workflows with advanced visualization tools that make it easy to compare multimodal outputs, analyze experiments, and uncover insights that are often missed in manual processes. In this video, you will learn how to train and analyze robotics models using W&B Models with NVIDIA Isaac Lab and Isaac Sim. Through a reinforcement learning project for a walking humanoid robot, we demonstrate how to tune key hyperparameters like learning rate, monitor training progress in real time, and scale workloads on H100 GPUs running on CoreWeave. You will also explore practical techniques for managing experiments at scale, such as pinning important runs, setting baselines, and organizing results so your most relevant insights are always within reach. See how tools like parallel coordinate plots help you understand the impact of hyperparameters on outcomes like reward and training stability.