<![CDATA[Bug_makerACE的博客]]>

<![CDATA[Bug_makerACE的博客]]>https://blog.youkuaiyun.com/Bug_makerACEzh-cnhttps://blog.youkuaiyun.com/<![CDATA[【深度长文】大模型分布式训练通信原语及其应用]]>https://blog.youkuaiyun.com/Bug_makerACE/article/details/158774371https://blog.youkuaiyun.com/Bug_makerACE/article/details/158774371Bug_makerACESat, 07 Mar 2026 17:19:43 +0800<![CDATA[飞书批量下载工具开发记录：从痛点到解决方案]]>https://blog.youkuaiyun.com/Bug_makerACE/article/details/158699246https://blog.youkuaiyun.com/Bug_makerACE/article/details/158699246Bug_makerACEThu, 05 Mar 2026 16:36:49 +0800<![CDATA[Elastic Attention：将 MoE 路由引入 Attention，让大模型长文本推理“该省省，该花花”]]>https://blog.youkuaiyun.com/Bug_makerACE/article/details/157434557https://blog.youkuaiyun.com/Bug_makerACE/article/details/157434557Bug_makerACETue, 27 Jan 2026 16:36:44 +0800<![CDATA[【深度长文】RLHF 算法全景解析：从 PPO 到 DPO/GRPO 的范式革命]]>https://blog.youkuaiyun.com/Bug_makerACE/article/details/155199475https://blog.youkuaiyun.com/Bug_makerACE/article/details/155199475Bug_makerACEMon, 24 Nov 2025 18:00:24 +0800<![CDATA[FP32/FP16/BF16/FP8...一文搞懂大模型数据类型变迁]]>https://blog.youkuaiyun.com/Bug_makerACE/article/details/154498611https://blog.youkuaiyun.com/Bug_makerACE/article/details/154498611Bug_makerACEThu, 06 Nov 2025 20:45:15 +0800<![CDATA[解读DeepSeek-V3.2-Exp：基于MLA架构的Lightning Index如何重塑长上下文效率]]>https://blog.youkuaiyun.com/Bug_makerACE/article/details/152522990https://blog.youkuaiyun.com/Bug_makerACE/article/details/152522990Bug_makerACESat, 04 Oct 2025 23:58:02 +0800<![CDATA[Stanford CS336 | Assignment 2 - FlashAttention-v2 Pytorch & Triotn实现]]>https://blog.youkuaiyun.com/Bug_makerACE/article/details/151656293https://blog.youkuaiyun.com/Bug_makerACE/article/details/151656293Bug_makerACESat, 13 Sep 2025 19:02:50 +0800<![CDATA[Stanford CS336 | Assignment 1 - Transformer Language Model Architecture]]>https://blog.youkuaiyun.com/Bug_makerACE/article/details/151655954https://blog.youkuaiyun.com/Bug_makerACE/article/details/151655954Bug_makerACESat, 13 Sep 2025 18:36:50 +0800<![CDATA[Standford CS336 | Assignment 1 - Implementing the tokenizer]]>https://blog.youkuaiyun.com/Bug_makerACE/article/details/149278120https://blog.youkuaiyun.com/Bug_makerACE/article/details/149278120Bug_makerACEFri, 11 Jul 2025 17:01:07 +0800<![CDATA[Stanford CS336 | Assignment 1 - BPE Tokenizer Training 实现]]>https://blog.youkuaiyun.com/Bug_makerACE/article/details/149248369https://blog.youkuaiyun.com/Bug_makerACE/article/details/149248369Bug_makerACEThu, 10 Jul 2025 14:07:19 +0800<![CDATA[ios::sync_with_stdio(false)和puts(“0“)会导致程序错误？谈C++输入输出同步的那些坑]]>https://blog.youkuaiyun.com/Bug_makerACE/article/details/147126125https://blog.youkuaiyun.com/Bug_makerACE/article/details/147126125Bug_makerACEThu, 10 Apr 2025 19:48:21 +0800<![CDATA[大模型（LLM）微调方法分类框架]]>https://blog.youkuaiyun.com/Bug_makerACE/article/details/146141616https://blog.youkuaiyun.com/Bug_makerACE/article/details/146141616Bug_makerACEMon, 10 Mar 2025 08:00:00 +0800<![CDATA[大模型解码策略详解：top-k、top-p与temperature]]>https://blog.youkuaiyun.com/Bug_makerACE/article/details/145923880https://blog.youkuaiyun.com/Bug_makerACE/article/details/145923880Bug_makerACEFri, 28 Feb 2025 11:31:17 +0800<![CDATA[LLaMA论文+源码分享]]>https://blog.youkuaiyun.com/Bug_makerACE/article/details/145629882https://blog.youkuaiyun.com/Bug_makerACE/article/details/145629882Bug_makerACEFri, 14 Feb 2025 11:50:48 +0800<![CDATA[一杯咖啡的时间学习大模型（LLM）：LLaMA解读之分组查询注意力（Grouped Query Attention）]]>https://blog.youkuaiyun.com/Bug_makerACE/article/details/145629111https://blog.youkuaiyun.com/Bug_makerACE/article/details/145629111Bug_makerACEFri, 14 Feb 2025 11:31:12 +0800<![CDATA[一杯咖啡的时间学习大模型（LLM）：LLaMA解读之SwiGLU激活函数]]>https://blog.youkuaiyun.com/Bug_makerACE/article/details/145626525https://blog.youkuaiyun.com/Bug_makerACE/article/details/145626525Bug_makerACEFri, 14 Feb 2025 09:57:20 +0800<![CDATA[一杯咖啡的时间学习大模型（LLM）：LLaMA解读之均方根误差标准化RMSNorm（含代码实现）]]>https://blog.youkuaiyun.com/Bug_makerACE/article/details/145621694https://blog.youkuaiyun.com/Bug_makerACE/article/details/145621694Bug_makerACEThu, 13 Feb 2025 22:35:39 +0800<![CDATA[一杯咖啡的时间学习大模型（LLM）：LLaMA解读之旋转编码RoPE（含代码实现）]]>https://blog.youkuaiyun.com/Bug_makerACE/article/details/145614732https://blog.youkuaiyun.com/Bug_makerACE/article/details/145614732Bug_makerACEThu, 13 Feb 2025 22:01:09 +0800<![CDATA[大模型（LLM）入门学习路线分享（内含计划表和时间安排）]]>https://blog.youkuaiyun.com/Bug_makerACE/article/details/145524603https://blog.youkuaiyun.com/Bug_makerACE/article/details/145524603Bug_makerACESat, 08 Feb 2025 23:16:33 +0800<![CDATA[【Linux基础】关于Linux Shell中参数的引用（双引号““、反斜杠）的疑难杂症]]>https://blog.youkuaiyun.com/Bug_makerACE/article/details/144047250https://blog.youkuaiyun.com/Bug_makerACE/article/details/144047250Bug_makerACETue, 26 Nov 2024 02:30:51 +0800