Day 3 – April 24, 2026
Breakouts
BRK1-070 • Compute • Introductory
Accelerate large-scale model pretraining and reinforcement learning fine-tuning
location_on
Breakers E
schedule
1:45 PM - 2:30 PM
Join this session to dive deep into the engineering behind training large, dense models for complex tasks like mathematical reasoning. We’ll take you through a specific case study of an end-to-end pipeline, highlighting the gotchas of convergence and tuning at this scale. Bring your questions and challenges for an interactive discussion on optimizing training loops, reducing computational overhead, and the future of compact model architectures.
Read more