Skip to content
AI HOT (Curated Pool)

Xiaohongshu and Peking University open-source UltraEP for real-time MoE load balancing

小红书与北大开源 UltraEP:面向大规模 MoE 训推的实时负载均衡方案

Xiaohongshu and Peking University open-sourced UltraEP, a real-time load balancing method for large MoE models. It dynamically replicates hot experts per microbatch and per layer using exact routing info, hitting 94.6% of ideal training throughput. On Qwen3-235B, training throughput is 42% higher than Megatron-LM and prefill throughput is 1.56× SGLang. The post doesn't disclose the license or deployment requirements.

Why it matters: Xiaohongshu and Peking University open-sourced UltraEP, which balances MoE load by dynamically replicating hot experts per layer and microbatch.

Read the original ↗Export Markdown