zzszmyf

zzszmyfhttps://zzszmyf.github.io/Recent content on zzszmyfHugozh-CN© 2026 zzszmyfMon, 17 Aug 2026 00:00:00 +0800LLM 推理优化 · 量化（Quantization）系统学习笔记https://zzszmyf.github.io/notes/llm%E6%8E%A8%E7%90%86%E4%BC%98%E5%8C%96-quantization%E9%87%8F%E5%8C%96%E5%AD%A6%E4%B9%A0%E7%AC%94%E8%AE%B0/Mon, 17 Aug 2026 00:00:00 +0800https://zzszmyf.github.io/notes/llm%E6%8E%A8%E7%90%86%E4%BC%98%E5%8C%96-quantization%E9%87%8F%E5%8C%96%E5%AD%A6%E4%B9%A0%E7%AC%94%E8%AE%B0/LLM 量化精读笔记 · 00 总览与学习地图https://zzszmyf.github.io/notes/llm%E9%87%8F%E5%8C%96%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-00-%E6%80%BB%E8%A7%88%E4%B8%8E%E5%AD%A6%E4%B9%A0%E5%9C%B0%E5%9B%BE/Mon, 17 Aug 2026 00:00:00 +0800https://zzszmyf.github.io/notes/llm%E9%87%8F%E5%8C%96%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-00-%E6%80%BB%E8%A7%88%E4%B8%8E%E5%AD%A6%E4%B9%A0%E5%9C%B0%E5%9B%BE/LLM 量化精读笔记 · 01 数值编码与计算机表示基础（信息论 · 整数 · IEEE 754 · 舍入与截断）https://zzszmyf.github.io/notes/llm%E9%87%8F%E5%8C%96%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-01-%E6%95%B0%E5%80%BC%E7%BC%96%E7%A0%81%E4%B8%8E%E8%AE%A1%E7%AE%97%E6%9C%BA%E8%A1%A8%E7%A4%BA%E5%9F%BA%E7%A1%80/Mon, 17 Aug 2026 00:00:00 +0800https://zzszmyf.github.io/notes/llm%E9%87%8F%E5%8C%96%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-01-%E6%95%B0%E5%80%BC%E7%BC%96%E7%A0%81%E4%B8%8E%E8%AE%A1%E7%AE%97%E6%9C%BA%E8%A1%A8%E7%A4%BA%E5%9F%BA%E7%A1%80/LLM 量化精读笔记 · 02 量化问题的形式化与均匀量化理论https://zzszmyf.github.io/notes/llm%E9%87%8F%E5%8C%96%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-02-%E9%87%8F%E5%8C%96%E9%97%AE%E9%A2%98%E5%BD%A2%E5%BC%8F%E5%8C%96%E4%B8%8E%E5%9D%87%E5%8C%80%E9%87%8F%E5%8C%96%E7%90%86%E8%AE%BA/Mon, 17 Aug 2026 00:00:00 +0800https://zzszmyf.github.io/notes/llm%E9%87%8F%E5%8C%96%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-02-%E9%87%8F%E5%8C%96%E9%97%AE%E9%A2%98%E5%BD%A2%E5%BC%8F%E5%8C%96%E4%B8%8E%E5%9D%87%E5%8C%80%E9%87%8F%E5%8C%96%E7%90%86%E8%AE%BA/LLM 量化精读笔记 · 03 数值格式：FP16 / BF16 / FP8 / FP4 / MXFP8 / NVFP4 与硬件https://zzszmyf.github.io/notes/llm%E9%87%8F%E5%8C%96%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-03-%E6%95%B0%E5%80%BC%E6%A0%BC%E5%BC%8F%E4%B8%8E%E7%A1%AC%E4%BB%B6/Mon, 17 Aug 2026 00:00:00 +0800https://zzszmyf.github.io/notes/llm%E9%87%8F%E5%8C%96%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-03-%E6%95%B0%E5%80%BC%E6%A0%BC%E5%BC%8F%E4%B8%8E%E7%A1%AC%E4%BB%B6/LLM 量化精读笔记 · 04 量化粒度、校准与离群值https://zzszmyf.github.io/notes/llm%E9%87%8F%E5%8C%96%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-04-%E9%87%8F%E5%8C%96%E7%B2%92%E5%BA%A6%E6%A0%A1%E5%87%86%E4%B8%8E%E7%A6%BB%E7%BE%A4%E5%80%BC/Mon, 17 Aug 2026 00:00:00 +0800https://zzszmyf.github.io/notes/llm%E9%87%8F%E5%8C%96%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-04-%E9%87%8F%E5%8C%96%E7%B2%92%E5%BA%A6%E6%A0%A1%E5%87%86%E4%B8%8E%E7%A6%BB%E7%BE%A4%E5%80%BC/LLM 量化精读笔记 · 05 权重量化 I：RTN 与 GPTQ（含二阶误差补偿推导）https://zzszmyf.github.io/notes/llm%E9%87%8F%E5%8C%96%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-05-%E6%9D%83%E9%87%8D%E9%87%8F%E5%8C%96i-rtn%E4%B8%8Egptq/Mon, 17 Aug 2026 00:00:00 +0800https://zzszmyf.github.io/notes/llm%E9%87%8F%E5%8C%96%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-05-%E6%9D%83%E9%87%8D%E9%87%8F%E5%8C%96i-rtn%E4%B8%8Egptq/LLM 量化精读笔记 · 06 权重量化 II：AWQ、SqueezeLLM、QuIP#https://zzszmyf.github.io/notes/llm%E9%87%8F%E5%8C%96%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-06-%E6%9D%83%E9%87%8D%E9%87%8F%E5%8C%96ii-awq-squeezellm-quip/Mon, 17 Aug 2026 00:00:00 +0800https://zzszmyf.github.io/notes/llm%E9%87%8F%E5%8C%96%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-06-%E6%9D%83%E9%87%8D%E9%87%8F%E5%8C%96ii-awq-squeezellm-quip/LLM 量化精读笔记 · 07 激活量化：LLM.int8() 与 SmoothQuant（W8A8）https://zzszmyf.github.io/notes/llm%E9%87%8F%E5%8C%96%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-07-%E6%BF%80%E6%B4%BB%E9%87%8F%E5%8C%96-llm-int8%E4%B8%8Esmoothquant/Mon, 17 Aug 2026 00:00:00 +0800https://zzszmyf.github.io/notes/llm%E9%87%8F%E5%8C%96%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-07-%E6%BF%80%E6%B4%BB%E9%87%8F%E5%8C%96-llm-int8%E4%B8%8Esmoothquant/LLM 量化精读笔记 · 08 KV Cache 量化：KIVI 与误差累积https://zzszmyf.github.io/notes/llm%E9%87%8F%E5%8C%96%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-08-kv-cache%E9%87%8F%E5%8C%96%E4%B8%8Ekivi/Mon, 17 Aug 2026 00:00:00 +0800https://zzszmyf.github.io/notes/llm%E9%87%8F%E5%8C%96%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-08-kv-cache%E9%87%8F%E5%8C%96%E4%B8%8Ekivi/LLM 量化精读笔记 · 09 QAT 与训练内量化：STE、QLoRA、BitNet b1.58https://zzszmyf.github.io/notes/llm%E9%87%8F%E5%8C%96%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-09-qat%E4%B8%8E%E8%AE%AD%E7%BB%83%E5%86%85%E9%87%8F%E5%8C%96-ste-qlora-bitnet/Mon, 17 Aug 2026 00:00:00 +0800https://zzszmyf.github.io/notes/llm%E9%87%8F%E5%8C%96%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-09-qat%E4%B8%8E%E8%AE%AD%E7%BB%83%E5%86%85%E9%87%8F%E5%8C%96-ste-qlora-bitnet/LLM 量化精读笔记 · 10 质量评估方法论：perplexity、基准与自定义评测https://zzszmyf.github.io/notes/llm%E9%87%8F%E5%8C%96%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-10-%E8%B4%A8%E9%87%8F%E8%AF%84%E4%BC%B0%E6%96%B9%E6%B3%95%E8%AE%BA/Mon, 17 Aug 2026 00:00:00 +0800https://zzszmyf.github.io/notes/llm%E9%87%8F%E5%8C%96%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-10-%E8%B4%A8%E9%87%8F%E8%AF%84%E4%BC%B0%E6%96%B9%E6%B3%95%E8%AE%BA/LLM 量化精读笔记 · 11 系统协同与部署：QServe、FP8 Attention 与工程工具https://zzszmyf.github.io/notes/llm%E9%87%8F%E5%8C%96%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-11-%E7%B3%BB%E7%BB%9F%E5%8D%8F%E5%90%8C%E4%B8%8E%E9%83%A8%E7%BD%B2/Mon, 17 Aug 2026 00:00:00 +0800https://zzszmyf.github.io/notes/llm%E9%87%8F%E5%8C%96%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-11-%E7%B3%BB%E7%BB%9F%E5%8D%8F%E5%90%8C%E4%B8%8E%E9%83%A8%E7%BD%B2/LLM 推测解码精读笔记 · 00 总览与学习地图https://zzszmyf.github.io/notes/llm%E6%8E%A8%E6%B5%8B%E8%A7%A3%E7%A0%81%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-00-%E6%80%BB%E8%A7%88%E4%B8%8E%E5%AD%A6%E4%B9%A0%E5%9C%B0%E5%9B%BE/Mon, 17 Aug 2026 00:00:00 +0800https://zzszmyf.github.io/notes/llm%E6%8E%A8%E6%B5%8B%E8%A7%A3%E7%A0%81%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-00-%E6%80%BB%E8%A7%88%E4%B8%8E%E5%AD%A6%E4%B9%A0%E5%9C%B0%E5%9B%BE/LLM 推测解码精读笔记 · 01 问题形式化与接受率数学https://zzszmyf.github.io/notes/llm%E6%8E%A8%E6%B5%8B%E8%A7%A3%E7%A0%81%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-01-%E9%97%AE%E9%A2%98%E5%BD%A2%E5%BC%8F%E5%8C%96%E4%B8%8E%E6%8E%A5%E5%8F%97%E7%8E%87%E6%95%B0%E5%AD%A6/Mon, 17 Aug 2026 00:00:00 +0800https://zzszmyf.github.io/notes/llm%E6%8E%A8%E6%B5%8B%E8%A7%A3%E7%A0%81%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-01-%E9%97%AE%E9%A2%98%E5%BD%A2%E5%BC%8F%E5%8C%96%E4%B8%8E%E6%8E%A5%E5%8F%97%E7%8E%87%E6%95%B0%E5%AD%A6/LLM 推测解码精读笔记 · 02 原始推测解码：草稿模型与拒绝采样https://zzszmyf.github.io/notes/llm%E6%8E%A8%E6%B5%8B%E8%A7%A3%E7%A0%81%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-02-%E5%8E%9F%E5%A7%8B%E6%8E%A8%E6%B5%8B%E8%A7%A3%E7%A0%81-%E8%8D%89%E7%A8%BF%E6%A8%A1%E5%9E%8B%E4%B8%8E%E6%8B%92%E7%BB%9D%E9%87%87%E6%A0%B7/Mon, 17 Aug 2026 00:00:00 +0800https://zzszmyf.github.io/notes/llm%E6%8E%A8%E6%B5%8B%E8%A7%A3%E7%A0%81%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-02-%E5%8E%9F%E5%A7%8B%E6%8E%A8%E6%B5%8B%E8%A7%A3%E7%A0%81-%E8%8D%89%E7%A8%BF%E6%A8%A1%E5%9E%8B%E4%B8%8E%E6%8B%92%E7%BB%9D%E9%87%87%E6%A0%B7/LLM 推测解码精读笔记 · 03 Medusa：多头解码https://zzszmyf.github.io/notes/llm%E6%8E%A8%E6%B5%8B%E8%A7%A3%E7%A0%81%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-03-medusa-%E5%A4%9A%E5%A4%B4%E8%A7%A3%E7%A0%81/Mon, 17 Aug 2026 00:00:00 +0800https://zzszmyf.github.io/notes/llm%E6%8E%A8%E6%B5%8B%E8%A7%A3%E7%A0%81%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-03-medusa-%E5%A4%9A%E5%A4%B4%E8%A7%A3%E7%A0%81/LLM 推测解码精读笔记 · 04 EAGLE：特征空间草稿https://zzszmyf.github.io/notes/llm%E6%8E%A8%E6%B5%8B%E8%A7%A3%E7%A0%81%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-04-eagle-%E7%89%B9%E5%BE%81%E7%A9%BA%E9%97%B4%E8%8D%89%E7%A8%BF/Mon, 17 Aug 2026 00:00:00 +0800https://zzszmyf.github.io/notes/llm%E6%8E%A8%E6%B5%8B%E8%A7%A3%E7%A0%81%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-04-eagle-%E7%89%B9%E5%BE%81%E7%A9%BA%E9%97%B4%E8%8D%89%E7%A8%BF/LLM 推测解码精读笔记 · 05 n-gram / 检索式与无模型路线https://zzszmyf.github.io/notes/llm%E6%8E%A8%E6%B5%8B%E8%A7%A3%E7%A0%81%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-05-n-gram%E6%A3%80%E7%B4%A2%E5%BC%8F%E4%B8%8E%E6%97%A0%E6%A8%A1%E5%9E%8B%E8%B7%AF%E7%BA%BF/Mon, 17 Aug 2026 00:00:00 +0800https://zzszmyf.github.io/notes/llm%E6%8E%A8%E6%B5%8B%E8%A7%A3%E7%A0%81%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-05-n-gram%E6%A3%80%E7%B4%A2%E5%BC%8F%E4%B8%8E%E6%97%A0%E6%A8%A1%E5%9E%8B%E8%B7%AF%E7%BA%BF/LLM 推测解码精读笔记 · 06 系统集成与生产验收https://zzszmyf.github.io/notes/llm%E6%8E%A8%E6%B5%8B%E8%A7%A3%E7%A0%81%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-06-%E7%B3%BB%E7%BB%9F%E9%9B%86%E6%88%90%E4%B8%8E%E7%94%9F%E4%BA%A7%E9%AA%8C%E6%94%B6/Mon, 17 Aug 2026 00:00:00 +0800https://zzszmyf.github.io/notes/llm%E6%8E%A8%E6%B5%8B%E8%A7%A3%E7%A0%81%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-06-%E7%B3%BB%E7%BB%9F%E9%9B%86%E6%88%90%E4%B8%8E%E7%94%9F%E4%BA%A7%E9%AA%8C%E6%94%B6/LLM 注意力与计算内核精读笔记 · 00 总览与学习地图https://zzszmyf.github.io/notes/llm%E6%B3%A8%E6%84%8F%E5%8A%9B%E5%86%85%E6%A0%B8%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-00-%E6%80%BB%E8%A7%88%E4%B8%8E%E5%AD%A6%E4%B9%A0%E5%9C%B0%E5%9B%BE/Mon, 17 Aug 2026 00:00:00 +0800https://zzszmyf.github.io/notes/llm%E6%B3%A8%E6%84%8F%E5%8A%9B%E5%86%85%E6%A0%B8%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-00-%E6%80%BB%E8%A7%88%E4%B8%8E%E5%AD%A6%E4%B9%A0%E5%9C%B0%E5%9B%BE/LLM 注意力与计算内核精读笔记 · 01 注意力机制基础与复杂度分析https://zzszmyf.github.io/notes/llm%E6%B3%A8%E6%84%8F%E5%8A%9B%E5%86%85%E6%A0%B8%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-01-%E6%B3%A8%E6%84%8F%E5%8A%9B%E6%9C%BA%E5%88%B6%E5%9F%BA%E7%A1%80%E4%B8%8E%E5%A4%8D%E6%9D%82%E5%BA%A6%E5%88%86%E6%9E%90/Mon, 17 Aug 2026 00:00:00 +0800https://zzszmyf.github.io/notes/llm%E6%B3%A8%E6%84%8F%E5%8A%9B%E5%86%85%E6%A0%B8%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-01-%E6%B3%A8%E6%84%8F%E5%8A%9B%E6%9C%BA%E5%88%B6%E5%9F%BA%E7%A1%80%E4%B8%8E%E5%A4%8D%E6%9D%82%E5%BA%A6%E5%88%86%E6%9E%90/LLM 注意力与计算内核精读笔记 · 02 FlashAttention：IO 感知的精确注意力https://zzszmyf.github.io/notes/llm%E6%B3%A8%E6%84%8F%E5%8A%9B%E5%86%85%E6%A0%B8%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-02-flashattention-io%E6%84%9F%E7%9F%A5%E7%9A%84%E7%B2%BE%E7%A1%AE%E6%B3%A8%E6%84%8F%E5%8A%9B/Mon, 17 Aug 2026 00:00:00 +0800https://zzszmyf.github.io/notes/llm%E6%B3%A8%E6%84%8F%E5%8A%9B%E5%86%85%E6%A0%B8%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-02-flashattention-io%E6%84%9F%E7%9F%A5%E7%9A%84%E7%B2%BE%E7%A1%AE%E6%B3%A8%E6%84%8F%E5%8A%9B/LLM 注意力与计算内核精读笔记 · 03 注意力头变体：MQA / GQA / MLAhttps://zzszmyf.github.io/notes/llm%E6%B3%A8%E6%84%8F%E5%8A%9B%E5%86%85%E6%A0%B8%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-03-%E6%B3%A8%E6%84%8F%E5%8A%9B%E5%A4%B4%E5%8F%98%E4%BD%93-mqa-gqa-mla/Mon, 17 Aug 2026 00:00:00 +0800https://zzszmyf.github.io/notes/llm%E6%B3%A8%E6%84%8F%E5%8A%9B%E5%86%85%E6%A0%B8%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-03-%E6%B3%A8%E6%84%8F%E5%8A%9B%E5%A4%B4%E5%8F%98%E4%BD%93-mqa-gqa-mla/LLM 注意力与计算内核精读笔记 · 04 稀疏、滑动窗口与线性注意力https://zzszmyf.github.io/notes/llm%E6%B3%A8%E6%84%8F%E5%8A%9B%E5%86%85%E6%A0%B8%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-04-%E7%A8%80%E7%96%8F%E6%BB%91%E5%8A%A8%E7%AA%97%E5%8F%A3%E4%B8%8E%E7%BA%BF%E6%80%A7%E6%B3%A8%E6%84%8F%E5%8A%9B/Mon, 17 Aug 2026 00:00:00 +0800https://zzszmyf.github.io/notes/llm%E6%B3%A8%E6%84%8F%E5%8A%9B%E5%86%85%E6%A0%B8%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-04-%E7%A8%80%E7%96%8F%E6%BB%91%E5%8A%A8%E7%AA%97%E5%8F%A3%E4%B8%8E%E7%BA%BF%E6%80%A7%E6%B3%A8%E6%84%8F%E5%8A%9B/LLM 注意力与计算内核精读笔记 · 05 PagedAttention 与 KV 显存管理https://zzszmyf.github.io/notes/llm%E6%B3%A8%E6%84%8F%E5%8A%9B%E5%86%85%E6%A0%B8%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-05-pagedattention%E4%B8%8Ekv%E6%98%BE%E5%AD%98%E7%AE%A1%E7%90%86/Mon, 17 Aug 2026 00:00:00 +0800https://zzszmyf.github.io/notes/llm%E6%B3%A8%E6%84%8F%E5%8A%9B%E5%86%85%E6%A0%B8%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-05-pagedattention%E4%B8%8Ekv%E6%98%BE%E5%AD%98%E7%AE%A1%E7%90%86/LLM 注意力与计算内核精读笔记 · 06 内核优化与算子融合https://zzszmyf.github.io/notes/llm%E6%B3%A8%E6%84%8F%E5%8A%9B%E5%86%85%E6%A0%B8%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-06-%E5%86%85%E6%A0%B8%E4%BC%98%E5%8C%96%E4%B8%8E%E7%AE%97%E5%AD%90%E8%9E%8D%E5%90%88/Mon, 17 Aug 2026 00:00:00 +0800https://zzszmyf.github.io/notes/llm%E6%B3%A8%E6%84%8F%E5%8A%9B%E5%86%85%E6%A0%B8%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-06-%E5%86%85%E6%A0%B8%E4%BC%98%E5%8C%96%E4%B8%8E%E7%AE%97%E5%AD%90%E8%9E%8D%E5%90%88/LLM 注意力与计算内核精读笔记 · 07 系统集成与生产验收https://zzszmyf.github.io/notes/llm%E6%B3%A8%E6%84%8F%E5%8A%9B%E5%86%85%E6%A0%B8%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-07-%E7%B3%BB%E7%BB%9F%E9%9B%86%E6%88%90%E4%B8%8E%E7%94%9F%E4%BA%A7%E9%AA%8C%E6%94%B6/Mon, 17 Aug 2026 00:00:00 +0800https://zzszmyf.github.io/notes/llm%E6%B3%A8%E6%84%8F%E5%8A%9B%E5%86%85%E6%A0%B8%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-07-%E7%B3%BB%E7%BB%9F%E9%9B%86%E6%88%90%E4%B8%8E%E7%94%9F%E4%BA%A7%E9%AA%8C%E6%94%B6/LLM 注意力与计算内核精读笔记 · 08 前缀缓存与 KV 复用https://zzszmyf.github.io/notes/llm%E6%B3%A8%E6%84%8F%E5%8A%9B%E5%86%85%E6%A0%B8%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-08-%E5%89%8D%E7%BC%80%E7%BC%93%E5%AD%98%E4%B8%8Ekv%E5%A4%8D%E7%94%A8/Mon, 17 Aug 2026 00:00:00 +0800https://zzszmyf.github.io/notes/llm%E6%B3%A8%E6%84%8F%E5%8A%9B%E5%86%85%E6%A0%B8%E7%B2%BE%E8%AF%BB%E7%AC%94%E8%AE%B0-08-%E5%89%8D%E7%BC%80%E7%BC%93%E5%AD%98%E4%B8%8Ekv%E5%A4%8D%E7%94%A8/AI Coding 的供给侧爆炸：需求侧何时跟上？https://zzszmyf.github.io/notes/ai-coding-supply-demand-explosion/Fri, 15 May 2026 10:40:00 +0800https://zzszmyf.github.io/notes/ai-coding-supply-demand-explosion/拖拽式简历编辑器：画布与拖拽技术选型指南https://zzszmyf.github.io/notes/drag-drop-canvas-resume-builder/Wed, 13 May 2026 10:00:00 +0800https://zzszmyf.github.io/notes/drag-drop-canvas-resume-builder/我们需要的不是钱，需要的是财富https://zzszmyf.github.io/notes/money-is-not-wealth/Sun, 10 May 2026 21:00:00 +0800https://zzszmyf.github.io/notes/money-is-not-wealth/Procfile、Nixpacks 与 Shebang：三个部署/脚本小知识https://zzszmyf.github.io/notes/procfile-nixpacks-shebang-notes/Fri, 08 May 2026 17:21:00 +0800https://zzszmyf.github.io/notes/procfile-nixpacks-shebang-notes/推荐系统中的 Hash 与位运算技巧：从 CityHash 到 Feature Hashinghttps://zzszmyf.github.io/notes/hash-bitwise-techniques-summary/Fri, 08 May 2026 15:15:00 +0800https://zzszmyf.github.io/notes/hash-bitwise-techniques-summary/vLLM GPU 利用率持续 100% 排查记：一个 max_tokens 参数引发的性能陷阱https://zzszmyf.github.io/notes/vllm-gpu-100-percent-max-tokens-trap/Fri, 08 May 2026 15:00:00 +0800https://zzszmyf.github.io/notes/vllm-gpu-100-percent-max-tokens-trap/React 组件库选型指南：Ant Design vs Material UI 及其他https://zzszmyf.github.io/notes/react-component-libraries-guide/Fri, 08 May 2026 10:00:00 +0800https://zzszmyf.github.io/notes/react-component-libraries-guide/用 kimi-cli 实现自动 Issue → Worktree → TDD → PR 的无人值守流水线https://zzszmyf.github.io/notes/kimi-cli-auto-issue-pipeline/Tue, 05 May 2026 14:00:00 +0800https://zzszmyf.github.io/notes/kimi-cli-auto-issue-pipeline/绩效评级看 Impact，晋升看 Behavior：职场晋升的隐藏规则https://zzszmyf.github.io/notes/%E7%BB%A9%E6%95%88%E8%AF%84%E7%BA%A7%E4%B8%8E%E6%99%8B%E5%8D%87%E7%9A%84%E5%8F%8C%E8%BD%A8%E8%AF%84%E4%BC%B0%E9%80%BB%E8%BE%91/Tue, 05 May 2026 10:16:00 +0800https://zzszmyf.github.io/notes/%E7%BB%A9%E6%95%88%E8%AF%84%E7%BA%A7%E4%B8%8E%E6%99%8B%E5%8D%87%E7%9A%84%E5%8F%8C%E8%BD%A8%E8%AF%84%E4%BC%B0%E9%80%BB%E8%BE%91/五一假期，我花了3.5亿Token，从零写了一套企业LLM WiKihttps://zzszmyf.github.io/notes/%E4%BA%94%E4%B8%80%E5%81%87%E6%9C%9F%E4%BB%8E%E9%9B%B6%E5%86%99%E4%BC%81%E4%B8%9Allm-wiki/Tue, 05 May 2026 08:00:00 +0800https://zzszmyf.github.io/notes/%E4%BA%94%E4%B8%80%E5%81%87%E6%9C%9F%E4%BB%8E%E9%9B%B6%E5%86%99%E4%BC%81%E4%B8%9Allm-wiki/塔勒布思想系列（6）：为什么我们听过很多道理，却依然过不好这一生？https://zzszmyf.github.io/notes/%E4%B8%BA%E4%BB%80%E4%B9%88%E6%88%91%E4%BB%AC%E5%90%AC%E8%BF%87%E5%BE%88%E5%A4%9A%E9%81%93%E7%90%86%E5%8D%B4%E4%BE%9D%E7%84%B6%E8%BF%87%E4%B8%8D%E5%A5%BD%E8%BF%99%E4%B8%80%E7%94%9F/Mon, 27 Apr 2026 08:38:00 +0800https://zzszmyf.github.io/notes/%E4%B8%BA%E4%BB%80%E4%B9%88%E6%88%91%E4%BB%AC%E5%90%AC%E8%BF%87%E5%BE%88%E5%A4%9A%E9%81%93%E7%90%86%E5%8D%B4%E4%BE%9D%E7%84%B6%E8%BF%87%E4%B8%8D%E5%A5%BD%E8%BF%99%E4%B8%80%E7%94%9F/工具型还是生命型：软件是该稳定一致，还是该成长变化https://zzszmyf.github.io/notes/living-software-%E5%B7%A5%E5%85%B7%E5%9E%8B%E4%B8%8E%E7%94%9F%E5%91%BD%E5%9E%8B/Mon, 27 Apr 2026 08:30:00 +0800https://zzszmyf.github.io/notes/living-software-%E5%B7%A5%E5%85%B7%E5%9E%8B%E4%B8%8E%E7%94%9F%E5%91%BD%E5%9E%8B/全员 token-maxxing，一场没人敢停的军备竞赛https://zzszmyf.github.io/notes/%E5%85%A8%E5%91%98-token-maxxing-%E5%86%9B%E5%A4%87%E7%AB%9E%E8%B5%9B/Mon, 27 Apr 2026 08:00:00 +0800https://zzszmyf.github.io/notes/%E5%85%A8%E5%91%98-token-maxxing-%E5%86%9B%E5%A4%87%E7%AB%9E%E8%B5%9B/这个博客是怎么做的：Hugo + Congo 完整技术拆解https://zzszmyf.github.io/notes/how-this-blog-is-built/Sat, 25 Jan 2025 10:00:00 +0800https://zzszmyf.github.io/notes/how-this-blog-is-built/Online Softmax 的数学：从代数推导到信息几何https://zzszmyf.github.io/notes/online-softmax-information-geometry/Mon, 20 Jan 2025 10:00:00 +0800https://zzszmyf.github.io/notes/online-softmax-information-geometry/逻辑回归权重更新：从交叉熵到梯度下降的完整推导https://zzszmyf.github.io/notes/logistic-regression-gradient-descent/Wed, 15 Jan 2025 10:00:00 +0800https://zzszmyf.github.io/notes/logistic-regression-gradient-descent/效率狂魔的自白：语音输入让我写作速度翻倍https://zzszmyf.github.io/notes/voice-input-writing-workflow/Mon, 16 Dec 2024 06:34:00 +0800https://zzszmyf.github.io/notes/voice-input-writing-workflow/VC-FCST Benchmark 落地方案：Code Agent 标准化测评https://zzszmyf.github.io/notes/vc-fcst-benchmark/Wed, 17 Apr 2024 00:00:00 +0800https://zzszmyf.github.io/notes/vc-fcst-benchmark/TurboQuant 数学原理解析https://zzszmyf.github.io/notes/turboquant-math-principles/Tue, 16 Apr 2024 00:00:00 +0800https://zzszmyf.github.io/notes/turboquant-math-principles/TurboQuant 学习路径：论文 → 技术 → 数学https://zzszmyf.github.io/notes/turboquant-learning-path/Mon, 15 Apr 2024 00:00:00 +0800https://zzszmyf.github.io/notes/turboquant-learning-path/进一步推进 MuP：从 Muon 到 SSOhttps://zzszmyf.github.io/notes/sso-spectral-sphere-optimization/Sun, 14 Apr 2024 00:00:00 +0800https://zzszmyf.github.io/notes/sso-spectral-sphere-optimization/From 10% to 96%: Qwen2.5-0.5B 文本分类微调实践https://zzszmyf.github.io/notes/qwen2.5-0.5b-text-classification-finetuning/Sat, 13 Apr 2024 00:00:00 +0800https://zzszmyf.github.io/notes/qwen2.5-0.5b-text-classification-finetuning/Puppeteer MCP 与 Playwright MCP：深度对比与选择指南https://zzszmyf.github.io/notes/puppeteer-vs-playwright-mcp/Fri, 12 Apr 2024 00:00:00 +0800https://zzszmyf.github.io/notes/puppeteer-vs-playwright-mcp/深入解读 Primus：面向大规模大语言模型的高性能训练框架https://zzszmyf.github.io/notes/primus-llm-training-framework/Thu, 11 Apr 2024 00:00:00 +0800https://zzszmyf.github.io/notes/primus-llm-training-framework/随便聊聊OpenClaw记忆系统https://zzszmyf.github.io/notes/openclaw-memory-system/Wed, 10 Apr 2024 00:00:00 +0800https://zzszmyf.github.io/notes/openclaw-memory-system/【万字硬核】微软 & NVIDIA 联合巨作《Using DeepSpeed and Megatron to Train Megatron-Turing NLG 530B》全方位技术解析https://zzszmyf.github.io/notes/mt-nlg-530b-deep-dive/Tue, 09 Apr 2024 00:00:00 +0800https://zzszmyf.github.io/notes/mt-nlg-530b-deep-dive/MoE数据特征、训练和推理的通信特点、基于NVLink/Scale Up那些Feature，DeepSeek的DeepEP和对MoE All-to-All数据的处理https://zzszmyf.github.io/notes/moe-nvlink-deepseek-deepep-communication/Mon, 08 Apr 2024 00:00:00 +0800https://zzszmyf.github.io/notes/moe-nvlink-deepseek-deepep-communication/LoRA 微调全流程实战——用 12 条数据、4 秒钟，把 Gemma 2B 训成特朗普https://zzszmyf.github.io/notes/lora-gemma2b-trump-finetuning/Sun, 07 Apr 2024 00:00:00 +0800https://zzszmyf.github.io/notes/lora-gemma2b-trump-finetuning/《KIMI K2.5: VISUAL AGENTIC INTELLIGENCE》阅读笔记https://zzszmyf.github.io/notes/kimi-k2-5-visual-agentic-intelligence/Sat, 06 Apr 2024 00:00:00 +0800https://zzszmyf.github.io/notes/kimi-k2-5-visual-agentic-intelligence/高维几何与集中不等式详解https://zzszmyf.github.io/notes/high-dim-geometry-concentration/Fri, 05 Apr 2024 00:00:00 +0800https://zzszmyf.github.io/notes/high-dim-geometry-concentration/Drifting Model vs Diffusion：复杂指令理解能力对比https://zzszmyf.github.io/notes/drifting_vs_diffusion/Thu, 04 Apr 2024 00:00:00 +0800https://zzszmyf.github.io/notes/drifting_vs_diffusion/百花齐放的 Diffusion-RL：技术路线综述https://zzszmyf.github.io/notes/diffusion-rl-methods/Wed, 03 Apr 2024 00:00:00 +0800https://zzszmyf.github.io/notes/diffusion-rl-methods/Claude Code Agent Teams 运行机制深度分析https://zzszmyf.github.io/notes/claude-code-agent-teams-deep-dive/Tue, 02 Apr 2024 00:00:00 +0800https://zzszmyf.github.io/notes/claude-code-agent-teams-deep-dive/从各家技术文章看26年Agentic RL Infra优化方向https://zzszmyf.github.io/notes/agentic-rl-infra-optimization-2026/Mon, 01 Apr 2024 00:00:00 +0800https://zzszmyf.github.io/notes/agentic-rl-infra-optimization-2026/完整技术栈推荐https://zzszmyf.github.io/notes/%E5%AE%8C%E6%95%B4%E6%8A%80%E6%9C%AF%E6%A0%88%E6%8E%A8%E8%8D%90/Sun, 24 Mar 2024 00:00:00 +0800https://zzszmyf.github.io/notes/%E5%AE%8C%E6%95%B4%E6%8A%80%E6%9C%AF%E6%A0%88%E6%8E%A8%E8%8D%90/后端部署方案对比（非静态网站）https://zzszmyf.github.io/notes/%E5%90%8E%E7%AB%AF%E9%83%A8%E7%BD%B2%E6%96%B9%E6%A1%88%E5%AF%B9%E6%AF%94/Sat, 23 Mar 2024 00:00:00 +0800https://zzszmyf.github.io/notes/%E5%90%8E%E7%AB%AF%E9%83%A8%E7%BD%B2%E6%96%B9%E6%A1%88%E5%AF%B9%E6%AF%94/Cloudflare 部署方案详解https://zzszmyf.github.io/notes/cloudflare%E9%83%A8%E7%BD%B2%E6%96%B9%E6%A1%88%E8%AF%A6%E8%A7%A3/Fri, 22 Mar 2024 00:00:00 +0800https://zzszmyf.github.io/notes/cloudflare%E9%83%A8%E7%BD%B2%E6%96%B9%E6%A1%88%E8%AF%A6%E8%A7%A3/海外域名提供商对比（.com）https://zzszmyf.github.io/notes/%E6%B5%B7%E5%A4%96%E5%9F%9F%E5%90%8D%E6%8F%90%E4%BE%9B%E5%95%86%E5%AF%B9%E6%AF%94/Thu, 21 Mar 2024 00:00:00 +0800https://zzszmyf.github.io/notes/%E6%B5%B7%E5%A4%96%E5%9F%9F%E5%90%8D%E6%8F%90%E4%BE%9B%E5%95%86%E5%AF%B9%E6%AF%94/域名购买与部署方案对比https://zzszmyf.github.io/notes/%E5%9F%9F%E5%90%8D%E8%B4%AD%E4%B9%B0%E4%B8%8E%E9%83%A8%E7%BD%B2%E6%96%B9%E6%A1%88%E5%AF%B9%E6%AF%94/Wed, 20 Mar 2024 00:00:00 +0800https://zzszmyf.github.io/notes/%E5%9F%9F%E5%90%8D%E8%B4%AD%E4%B9%B0%E4%B8%8E%E9%83%A8%E7%BD%B2%E6%96%B9%E6%A1%88%E5%AF%B9%E6%AF%94/域名、SEO 与 GEO 实战指南https://zzszmyf.github.io/notes/%E5%9F%9F%E5%90%8Dseo%E4%B8%8Egeo%E5%AE%9E%E6%88%98%E6%8C%87%E5%8D%97/Tue, 19 Mar 2024 00:00:00 +0800https://zzszmyf.github.io/notes/%E5%9F%9F%E5%90%8Dseo%E4%B8%8Egeo%E5%AE%9E%E6%88%98%E6%8C%87%E5%8D%97/推荐资源清单https://zzszmyf.github.io/notes/%E6%8E%A8%E8%8D%90%E8%B5%84%E6%BA%90%E6%B8%85%E5%8D%95/Mon, 18 Mar 2024 00:00:00 +0800https://zzszmyf.github.io/notes/%E6%8E%A8%E8%8D%90%E8%B5%84%E6%BA%90%E6%B8%85%E5%8D%95/90 天启动计划https://zzszmyf.github.io/notes/90%E5%A4%A9%E5%90%AF%E5%8A%A8%E8%AE%A1%E5%88%92/Sun, 17 Mar 2024 00:00:00 +0800https://zzszmyf.github.io/notes/90%E5%A4%A9%E5%90%AF%E5%8A%A8%E8%AE%A1%E5%88%92/常见陷阱与避坑指南https://zzszmyf.github.io/notes/%E5%B8%B8%E8%A7%81%E9%99%B7%E9%98%B1%E4%B8%8E%E9%81%BF%E5%9D%91%E6%8C%87%E5%8D%97/Sat, 16 Mar 2024 00:00:00 +0800https://zzszmyf.github.io/notes/%E5%B8%B8%E8%A7%81%E9%99%B7%E9%98%B1%E4%B8%8E%E9%81%BF%E5%9D%91%E6%8C%87%E5%8D%97/内容营销执行手册https://zzszmyf.github.io/notes/%E5%86%85%E5%AE%B9%E8%90%A5%E9%94%80%E6%89%A7%E8%A1%8C%E6%89%8B%E5%86%8C/Fri, 15 Mar 2024 00:00:00 +0800https://zzszmyf.github.io/notes/%E5%86%85%E5%AE%B9%E8%90%A5%E9%94%80%E6%89%A7%E8%A1%8C%E6%89%8B%E5%86%8C/垂直行业选择指南https://zzszmyf.github.io/notes/%E5%9E%82%E7%9B%B4%E8%A1%8C%E4%B8%9A%E9%80%89%E6%8B%A9%E6%8C%87%E5%8D%97/Thu, 14 Mar 2024 00:00:00 +0800https://zzszmyf.github.io/notes/%E5%9E%82%E7%9B%B4%E8%A1%8C%E4%B8%9A%E9%80%89%E6%8B%A9%E6%8C%87%E5%8D%97/顾问形象与心态建设https://zzszmyf.github.io/notes/%E9%A1%BE%E9%97%AE%E5%BD%A2%E8%B1%A1%E4%B8%8E%E5%BF%83%E6%80%81%E5%BB%BA%E8%AE%BE/Wed, 13 Mar 2024 00:00:00 +0800https://zzszmyf.github.io/notes/%E9%A1%BE%E9%97%AE%E5%BD%A2%E8%B1%A1%E4%B8%8E%E5%BF%83%E6%80%81%E5%BB%BA%E8%AE%BE/客户沟通与提案技巧https://zzszmyf.github.io/notes/%E5%AE%A2%E6%88%B7%E6%B2%9F%E9%80%9A%E4%B8%8E%E6%8F%90%E6%A1%88%E6%8A%80%E5%B7%A7/Tue, 12 Mar 2024 00:00:00 +0800https://zzszmyf.github.io/notes/%E5%AE%A2%E6%88%B7%E6%B2%9F%E9%80%9A%E4%B8%8E%E6%8F%90%E6%A1%88%E6%8A%80%E5%B7%A7/TOB 咨询商务准备清单https://zzszmyf.github.io/notes/tob%E5%92%A8%E8%AF%A2%E5%95%86%E5%8A%A1%E5%87%86%E5%A4%87%E6%B8%85%E5%8D%95/Mon, 11 Mar 2024 00:00:00 +0800https://zzszmyf.github.io/notes/tob%E5%92%A8%E8%AF%A2%E5%95%86%E5%8A%A1%E5%87%86%E5%A4%87%E6%B8%85%E5%8D%95/AI 落地专家内容营销策略https://zzszmyf.github.io/notes/ai%E8%90%BD%E5%9C%B0%E4%B8%93%E5%AE%B6%E5%86%85%E5%AE%B9%E8%90%A5%E9%94%80%E7%AD%96%E7%95%A5/Sun, 10 Mar 2024 00:00:00 +0800https://zzszmyf.github.io/notes/ai%E8%90%BD%E5%9C%B0%E4%B8%93%E5%AE%B6%E5%86%85%E5%AE%B9%E8%90%A5%E9%94%80%E7%AD%96%E7%95%A5/技术品牌建设指南https://zzszmyf.github.io/notes/%E6%8A%80%E6%9C%AF%E5%93%81%E7%89%8C%E5%BB%BA%E8%AE%BE%E6%8C%87%E5%8D%97/Sat, 09 Mar 2024 00:00:00 +0800https://zzszmyf.github.io/notes/%E6%8A%80%E6%9C%AF%E5%93%81%E7%89%8C%E5%BB%BA%E8%AE%BE%E6%8C%87%E5%8D%97/博客非技术准备清单https://zzszmyf.github.io/notes/%E5%8D%9A%E5%AE%A2%E9%9D%9E%E6%8A%80%E6%9C%AF%E5%87%86%E5%A4%87%E6%B8%85%E5%8D%95/Fri, 08 Mar 2024 00:00:00 +0800https://zzszmyf.github.io/notes/%E5%8D%9A%E5%AE%A2%E9%9D%9E%E6%8A%80%E6%9C%AF%E5%87%86%E5%A4%87%E6%B8%85%E5%8D%95/快速开始指南https://zzszmyf.github.io/notes/%E5%BF%AB%E9%80%9F%E5%BC%80%E5%A7%8B%E6%8C%87%E5%8D%97/Thu, 07 Mar 2024 00:00:00 +0800https://zzszmyf.github.io/notes/%E5%BF%AB%E9%80%9F%E5%BC%80%E5%A7%8B%E6%8C%87%E5%8D%97/关于https://zzszmyf.github.io/about/Thu, 07 Mar 2024 00:00:00 +0800https://zzszmyf.github.io/about/