Running 1.6k 1.6k The Ultra-Scale Playbook 🌌 The ultimate guide to training LLM on large GPU Clusters