Profiling on k4i's bloghttps://k4i.top/zh/tags/profiling/Recent content in Profiling on k4i's blogHugo -- gohugo.iozhsky_io@outlook.com (K4i)sky_io@outlook.com (K4i)All content is subject to the license of <a rel="license noopener" href="https://creativecommons.org/licenses/by-nc-sa/4.0/" target="_blank">CC BY-NC-SA 4.0</a> .Fri, 05 Jun 2026 10:00:00 +0800LLM Inference Lab Reports：推理实验与 Benchmark 路线https://k4i.top/zh/posts/llm-inference-lab-reports/Fri, 05 Jun 2026 10:00:00 +0800sky_io@outlook.com (K4i)Fri, 05 Jun 2026 00:26:17 +0800https://k4i.top/zh/posts/llm-inference-lab-reports/<p>这条系列专门放实验报告。它和源码阅读、机制解释的区别是：每篇文章都要有可复现环境、命令、指标、图表或表格，以及明确的调参结论。</p> <p>推理引擎岗位面试里，“知道 PagedAttention / prefix cache / chunked prefill”只是第一层。更有说服力的是能回答：在什么 workload 下它有效？指标改善多少？瓶颈从哪里转移到了哪里？如果线上指标变差，应该先看什么？</p>K4ifeatured imagellm推理benchmarkprofilingvllmsglangAILLM Inference Lab Reports