tracked author

Yanghua Peng

Research scientist and engineering lead at Seed Infra, ByteDance. Recurs across MegaScale, HybridFlow/VERL, MegaScale-MoE, and Laminar infrastructure lines. No high-confidence personal X account found in this pass.

3 archived notes X: not-found Homepage

Related Notes

按论文归档时间排序，展示该作者在本站已经出现的材料。

归档 Jun 28, 2026

Laminar: A Scalable Asynchronous RL Post Training Framework

让完成轨迹独立进入 experience buffer，并以 CPU/RDMA relay 和同版本 repack 解除全局 batch / 权重同步 barrier。

待审阅 2510.12633-laminar-asynchronous-rl-post-training RL Infrastructure Rollout Optimization Distributed Training

归档 Jun 28, 2026

MegaScale MoE: Large Scale Communication Efficient Training of Mixture of Experts Models in Production

围绕 attention SP、FFN EP、communication overlap 和 compressed communication 重构生产 MoE 训练路径。

待审阅 2505.11432-megascale-moe-communication-efficient-training MoE Systems Distributed Training Training Memory

归档 Jan 27, 2026

HybridFlow: A Flexible and Efficient RLHF Framework

用跨模型 single controller 与模型内 multi controller 统一编排 RLHF dataflow 和并行执行。

待审阅 2409.19256-hybridflow-rlhf-framework RL Infrastructure Distributed Training Rollout Optimization