https://mzf666.github.io/llm-infra/zh/tags/distributed-training/Recent content in Distributed-Training on LLM Infra TutorialHugozh-CNSun, 05 Apr 2026 00:00:00 +0000https://mzf666.github.io/llm-infra/zh/posts/02-parallel-strategies/Mon, 16 Mar 2026 00:00:00 +0000https://mzf666.github.io/llm-infra/zh/posts/02-parallel-strategies/从 DDP 到混合并行，系统梳理大模型训练中的所有并行策略。https://mzf666.github.io/llm-infra/zh/posts/04-rlhf-system/Tue, 17 Mar 2026 00:00:00 +0000https://mzf666.github.io/llm-infra/zh/posts/04-rlhf-system/从 RLHF 四模型架构到 verl 系统实现，理解为什么 RLHF 本质上是一个系统问题。https://mzf666.github.io/llm-infra/zh/posts/05-megatron-bridge/Sun, 05 Apr 2026 00:00:00 +0000https://mzf666.github.io/llm-infra/zh/posts/05-megatron-bridge/从 Megatron-Core 架构到 Megatron-Bridge 桥接机制，深入剖析 MoE 大模型训练中的并行策略、通信优化与计算融合。