<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>天庭</title><description>记录王勇顺的AI学习,工程实践,以及各种经历。</description><link>https://tiantingnote.com/</link><item><title>学习大模型前你需要了解的组件</title><link>https://tiantingnote.com/posts/llm_scripts/</link><guid isPermaLink="true">https://tiantingnote.com/posts/llm_scripts/</guid><description>汇总大模型大部分组件、概念</description><pubDate>Thu, 11 Jun 2026 00:00:00 GMT</pubDate></item><item><title>Tokenizer(一)，最简单的BPE算法构建的Tokenizer</title><link>https://tiantingnote.com/posts/tokenizer/</link><guid isPermaLink="true">https://tiantingnote.com/posts/tokenizer/</guid><description>汇总大模型大部分组件、概念</description><pubDate>Thu, 11 Jun 2026 00:00:00 GMT</pubDate></item><item><title>芯宝是怎么诞生的</title><link>https://tiantingnote.com/posts/xin_bao_creation/</link><guid isPermaLink="true">https://tiantingnote.com/posts/xin_bao_creation/</guid><description>一个桌面级对话机器人，具有情感陪伴和随身知识库功能</description><pubDate>Thu, 21 May 2026 00:00:00 GMT</pubDate></item><item><title>Transformer 架构中的数学美学与注意力机制推导</title><link>https://tiantingnote.com/posts/transformer-math-notes/</link><guid isPermaLink="true">https://tiantingnote.com/posts/transformer-math-notes/</guid><description>系统推导 Transformer 中 Scaled Dot-Product Attention 缩放因子的数学原理，说明为什么要除以根号 d_k。</description><pubDate>Mon, 18 May 2026 00:00:00 GMT</pubDate></item><item><title>LLM 训练踩坑笔记：从数据清洗到损失稳定</title><link>https://tiantingnote.com/posts/llm-training-notes/</link><guid isPermaLink="true">https://tiantingnote.com/posts/llm-training-notes/</guid><description>整理一次中文语料训练中的关键坑位，包括分词污染、重复样本和 loss 波动处理。</description><pubDate>Mon, 20 Apr 2026 00:00:00 GMT</pubDate></item><item><title>RAG 检索设计实践：召回、重排与延迟平衡</title><link>https://tiantingnote.com/posts/rag-retrieval-design/</link><guid isPermaLink="true">https://tiantingnote.com/posts/rag-retrieval-design/</guid><description>以一个问答系统为例，拆解向量召回和重排在效果与成本之间的权衡。</description><pubDate>Sat, 18 Apr 2026 00:00:00 GMT</pubDate></item></channel></rss>