Yihua Zhang

Yihua Zhang https://normaluhr.github.io/ en Sat, 22 Aug 2026 05:26:01 GMT 《Vogel im Käfig》音乐解构：泽野弘之如何用最小的音程构造希望与绝望 https://normaluhr.github.io/en/p/vogel-im-kafig https://normaluhr.github.io/en/p/vogel-im-kafig Fri, 13 Feb 2026 00:00:00 GMT A Role Shift for AI Infra: From Foundational Support to a Core Engine of Innovation https://normaluhr.github.io/en/p/ai-infra https://normaluhr.github.io/en/p/ai-infra Thu, 02 Oct 2025 00:00:00 GMT AI Market Insights From GRPO to DAPO and GSPO: What, Why and How https://normaluhr.github.io/en/p/grpo-dapo-gspo https://normaluhr.github.io/en/p/grpo-dapo-gspo Fri, 08 Aug 2025 00:00:00 GMT Re-understanding KL Approximation from an RL-for-LLM Lens: Notes on “Approximating KL Divergence https://normaluhr.github.io/en/p/kl https://normaluhr.github.io/en/p/kl Wed, 02 Jul 2025 00:00:00 GMT Decorators in Machine Learning Projects https://normaluhr.github.io/en/p/decorators https://normaluhr.github.io/en/p/decorators Thu, 10 Apr 2025 00:00:00 GMT 《Bauklötze》音乐解构 https://normaluhr.github.io/en/p/bauklotze https://normaluhr.github.io/en/p/bauklotze Sat, 08 Mar 2025 00:00:00 GMT 积木崩塌时的命运回响，泽野弘之用音符砌筑的巨人悲歌 DualPipe Explained: A Comprehensive Guide to DualPipe That Anyone Can Understand—Even Without a Distributed Background https://normaluhr.github.io/en/p/dualpipe https://normaluhr.github.io/en/p/dualpipe Thu, 27 Feb 2025 00:00:00 GMT Navigating the RLHF Landscape: From Policy Gradients to PPO, GAE, and DPO for LLM Alignment https://normaluhr.github.io/en/p/rlhf https://normaluhr.github.io/en/p/rlhf Tue, 11 Feb 2025 00:00:00 GMT DeepSeek-R1 Dissection: Understanding PPO & GRPO Without Any Prior Reinforcement Learning Knowledge https://normaluhr.github.io/en/p/grpo https://normaluhr.github.io/en/p/grpo Fri, 07 Feb 2025 00:00:00 GMT Why Cache 32 Heads When One Latent Variable Suffices? A Theory-to-Code Guide to DeepSeek’s MLA for KV-Cache https://normaluhr.github.io/en/p/mla https://normaluhr.github.io/en/p/mla Sun, 02 Feb 2025 00:00:00 GMT From Zero to Reasoning Hero: How DeepSeek-R1 Leverages Reinforcement Learning to Master Complex Reasoning https://normaluhr.github.io/en/p/deepseek-r1 https://normaluhr.github.io/en/p/deepseek-r1 Mon, 20 Jan 2025 00:00:00 GMT A Review on the Evolvement of Load Balancing Strategy in MoE LLMs: Pitfalls and Lessons https://normaluhr.github.io/en/p/moe-load-balancing https://normaluhr.github.io/en/p/moe-load-balancing Wed, 15 Jan 2025 00:00:00 GMT Patching the Foundation Models: Pitfalls and Pains in Machine Unlearning https://normaluhr.github.io/en/p/unlearning-pitfalls https://normaluhr.github.io/en/p/unlearning-pitfalls Sun, 15 Dec 2024 00:00:00 GMT Reading Notes - Talk about Effective C++ in My Own Words https://normaluhr.github.io/en/p/effective-c https://normaluhr.github.io/en/p/effective-c Thu, 31 Dec 2020 00:00:00 GMT 漫谈C++——C++17中的constexpr https://normaluhr.github.io/en/p/constexpr-4 https://normaluhr.github.io/en/p/constexpr-4 Wed, 04 Nov 2020 00:00:00 GMT C++学习笔记漫谈C++——const真理大讨论之 const的线程安全 https://normaluhr.github.io/en/p/const-3 https://normaluhr.github.io/en/p/const-3 Tue, 03 Nov 2020 00:00:00 GMT C++学习笔记漫谈C++——const真理大讨论之 mutable https://normaluhr.github.io/en/p/const-2 https://normaluhr.github.io/en/p/const-2 Mon, 02 Nov 2020 00:00:00 GMT C++学习笔记漫谈C++——const真理大讨论之语法和语义const https://normaluhr.github.io/en/p/const-1 https://normaluhr.github.io/en/p/const-1 Sun, 01 Nov 2020 00:00:00 GMT C++学习笔记漫谈C++——谈谈在函数中使用auto https://normaluhr.github.io/en/p/auto-function https://normaluhr.github.io/en/p/auto-function Fri, 30 Oct 2020 00:00:00 GMT C++学习笔记漫谈C++——工厂模式中的通行证策略 https://normaluhr.github.io/en/p/passkey-idiom https://normaluhr.github.io/en/p/passkey-idiom Tue, 27 Oct 2020 00:00:00 GMT C++学习笔记漫谈C++——动态与静态 https://normaluhr.github.io/en/p/dynamic https://normaluhr.github.io/en/p/dynamic Mon, 05 Oct 2020 00:00:00 GMT C++学习笔记 Visual Slam笔记——李群和李代数 https://normaluhr.github.io/en/p/lie-group https://normaluhr.github.io/en/p/lie-group Fri, 25 Sep 2020 00:00:00 GMT Visual SLAM笔记漫谈C++——谈谈在变量中使用auto https://normaluhr.github.io/en/p/auto https://normaluhr.github.io/en/p/auto Thu, 24 Sep 2020 00:00:00 GMT C++学习笔记漫谈C++——override和final让你的虚函数更安全 https://normaluhr.github.io/en/p/override-final https://normaluhr.github.io/en/p/override-final Wed, 23 Sep 2020 00:00:00 GMT C++学习笔记漫谈C++——从编译期常量到constexpr（三） https://normaluhr.github.io/en/p/constexpr-3 https://normaluhr.github.io/en/p/constexpr-3 Sun, 20 Sep 2020 00:00:00 GMT C++学习笔记漫谈C++——从编译期常量到constexpr（二） https://normaluhr.github.io/en/p/constexpr-2 https://normaluhr.github.io/en/p/constexpr-2 Tue, 15 Sep 2020 00:00:00 GMT C++学习笔记漫谈C++——从编译期常量到constexpr（一） https://normaluhr.github.io/en/p/constexpr-1 https://normaluhr.github.io/en/p/constexpr-1 Mon, 14 Sep 2020 00:00:00 GMT C++学习笔记漫谈C++——C++17新特性之std::optional https://normaluhr.github.io/en/p/optional https://normaluhr.github.io/en/p/optional Fri, 11 Sep 2020 00:00:00 GMT C++学习笔记漫谈C++——用了这么久，你真的懂nullptr吗 https://normaluhr.github.io/en/p/nullptr https://normaluhr.github.io/en/p/nullptr Thu, 03 Sep 2020 00:00:00 GMT C++学习笔记漫谈C++——起底万能指针void* https://normaluhr.github.io/en/p/void https://normaluhr.github.io/en/p/void Sat, 15 Aug 2020 00:00:00 GMT C++学习笔记漫谈C++——右值引用与移动语义 https://normaluhr.github.io/en/p/rvalue https://normaluhr.github.io/en/p/rvalue Sat, 01 Aug 2020 00:00:00 GMT C++学习笔记漫谈C++——强枚举到底有多强 https://normaluhr.github.io/en/p/enum https://normaluhr.github.io/en/p/enum Tue, 28 Jul 2020 00:00:00 GMT C++学习笔记漫谈C++——顶层和底层const https://normaluhr.github.io/en/p/top-const https://normaluhr.github.io/en/p/top-const Tue, 18 Feb 2020 00:00:00 GMT C++学习笔记