Archive

2026  2

May  2

Making Long-Context MoE RL Training Easier to Tune: Optimization Practice in Megatron-Lite / bumblebee

May 18, 2026 · 22 min · Yan Bai

FSDP, PP, CP, and EP: Four Parallel Dimensions in Large-Scale Training

May 17, 2026 · 1 min · Yan Bai