deploy: 69c2890aad

deploy: 16732daa57
deploy: 3ee20f1118
2025-08-20 06:29:20 +00:00 · 2025-08-20 06:24:28 +00:00 · 2025-08-20 06:05:30 +00:00 · 2025-08-20 06:04:04 +00:00 · 2025-08-20 04:50:28 +00:00 · 2025-08-20 04:34:01 +00:00
57 changed files with 767 additions and 14066 deletions
--- a/.drone.yml
+++ b/.drone.yml
@@ -1,28 +0,0 @@
-kind: pipeline
-name: default
-
-steps:
- name: build
-  image: plugins/hugo
-  settings:
-    hugo_version: 0.97.0
-    extended: true
-    minify: true
-    pull: always
-    url: ericxliu.me
-    validate: false
-    output: "./output"
-    mtu: 1450
- name: git-push
-  image: appleboy/drone-git-push:0.2.0-linux-amd64
-  settings:
-    branch: gitea-pages
-    remote: "git@git.ericxliu.me:eric/ericxliu-me.git"
-    force: true
-    commit: true
-    path: "./output"
-    commit_message: "Drone build ${DRONE_COMMIT_SHA:0:7}"
-    author_name: "Eric Liu"
-    author_email: "eric@ericxliu.me"
-    ssh_key:
-      from_secret: ssh_key
--- a/.gitignore
+++ b/.gitignore
@@ -1 +0,0 @@
-_gen/
--- a/.gitlab-ci.yml
+++ b/.gitlab-ci.yml
@@ -1,37 +0,0 @@
-# This file is a template, and might need editing before it works on your project.
-# To contribute improvements to CI/CD templates, please follow the Development guide at:
-# https://docs.gitlab.com/ee/development/cicd/templates.html
-# This specific template is located at:
-# https://gitlab.com/gitlab-org/gitlab/-/blob/master/lib/gitlab/ci/templates/Pages/HTML.gitlab-ci.yml
-
-# Full project: https://gitlab.com/pages/plain-html
-
-variables:
-  GIT_SUBMODULE_STRATEGY: recursive
-
-build-stage:
-  stage: build
-  image: monachus/hugo:latest
-  script:
-    - hugo
-    - ls 
-  artifacts:
-    paths:
-      - public
-
-deploy-stage:
-  stage: deploy
-  image: minio/mc:latest
-  script:
-    - ls 
-    - mkdir .public
-    - cp -r public/* .public
-    - mc alias set minio http://minio.diskstation.local:80 WjaYWk3uthUlotbT Hc3fff7v69nZ6XvcXXpOZ3JJMzcmGc6A
-    - mc cp -r .public/ minio/eric-personal
-  artifacts:
-    paths:
-      - .public
-  dependencies:
-    - build-stage
-  rules:
-    - if: $CI_COMMIT_BRANCH == $CI_DEFAULT_BRANCH
--- a/.gitmodules
+++ b/.gitmodules
@@ -1,3 +0,0 @@
-[submodule "themes/hugo-coder"]
-	path = themes/hugo-coder
-	url = https://github.com/luizdepra/hugo-coder
--- a/.nojekyll
+++ b/.nojekyll
--- a/404.html
+++ b/404.html
@@ -0,0 +1,7 @@
+<!doctype html><html lang=en><head><title>Eric X. Liu's Personal Page</title><meta charset=utf-8><meta name=viewport content="width=device-width,initial-scale=1"><meta name=color-scheme content="light dark"><meta name=author content="Eric X. Liu"><meta name=description content="Eric X. Liu - Software & Performance Engineer at Google. Sharing insights about software engineering, performance optimization, tech industry experiences, mountain biking adventures, Jeep overlanding, and outdoor activities."><meta name=keywords content="software engineer,performance engineering,Google engineer,tech blog,software development,performance optimization,Eric Liu,engineering blog,mountain biking,Jeep enthusiast,overlanding,camping,outdoor adventures"><meta name=twitter:card content="summary"><meta name=twitter:title content="404 Page not found"><meta name=twitter:description content="Eric X. Liu - Software & Performance Engineer at Google. Sharing insights about software engineering, performance optimization, tech industry experiences, mountain biking adventures, Jeep overlanding, and outdoor activities."><meta property="og:url" content="/404.html"><meta property="og:site_name" content="Eric X. Liu's Personal Page"><meta property="og:title" content="404 Page not found"><meta property="og:description" content="Eric X. Liu - Software & Performance Engineer at Google. Sharing insights about software engineering, performance optimization, tech industry experiences, mountain biking adventures, Jeep overlanding, and outdoor activities."><meta property="og:locale" content="en"><meta property="og:type" content="website"><link rel=canonical href=/404.html><link rel=preload href=/fonts/fa-brands-400.woff2 as=font type=font/woff2 crossorigin><link rel=preload href=/fonts/fa-regular-400.woff2 as=font type=font/woff2 crossorigin><link rel=preload href=/fonts/fa-solid-900.woff2 as=font type=font/woff2 crossorigin><link rel=stylesheet href=/css/coder.min.6445a802b9389c9660e1b07b724dcf5718b1065ed2d71b4eeaf981cc7cc5fc46.css integrity="sha256-ZEWoArk4nJZg4bB7ck3PVxixBl7S1xtO6vmBzHzF/EY=" crossorigin=anonymous media=screen><link rel=stylesheet href=/css/coder-dark.min.a00e6364bacbc8266ad1cc81230774a1397198f8cfb7bcba29b7d6fcb54ce57f.css integrity="sha256-oA5jZLrLyCZq0cyBIwd0oTlxmPjPt7y6KbfW/LVM5X8=" crossorigin=anonymous media=screen><link rel=icon type=image/svg+xml href=/images/favicon.svg sizes=any><link rel=icon type=image/png href=/images/favicon-32x32.png sizes=32x32><link rel=icon type=image/png href=/images/favicon-16x16.png sizes=16x16><link rel=apple-touch-icon href=/images/apple-touch-icon.png><link rel=apple-touch-icon sizes=180x180 href=/images/apple-touch-icon.png><link rel=manifest href=/site.webmanifest><link rel=mask-icon href=/images/safari-pinned-tab.svg color=#5bbad5></head><body class="preload-transitions colorscheme-auto"><div class=float-container><a id=dark-mode-toggle class=colorscheme-toggle><i class="fa-solid fa-adjust fa-fw" aria-hidden=true></i></a></div><main class=wrapper><nav class=navigation><section class=container><a class=navigation-title href=/>Eric X. Liu's Personal Page
+</a><input type=checkbox id=menu-toggle>
+<label class="menu-button float-right" for=menu-toggle><i class="fa-solid fa-bars fa-fw" aria-hidden=true></i></label><ul class=navigation-list><li class=navigation-item><a class=navigation-link href=/posts/>Posts</a></li><li class=navigation-item><a class=navigation-link href=https://chat.ericxliu.me>Chat</a></li><li class=navigation-item><a class=navigation-link href=https://git.ericxliu.me/user/oauth2/Authenitk>Git</a></li><li class=navigation-item><a class=navigation-link href=https://coder.ericxliu.me/api/v2/users/oidc/callback>Coder</a></li><li class=navigation-item><a class=navigation-link href=/>|</a></li><li class=navigation-item><a class=navigation-link href=https://sso.ericxliu.me>Sign in</a></li></ul></section></nav><div class=content><section class="container centered"><div class=error><h1>404</h1><h2>Page Not Found</h2><p>Sorry, this page does not exist.<br>You can head back to the <a href=/>homepage</a>.</p></div></section></div><footer class=footer><section class=container>©
+2016 -
+2025
+Eric X. Liu
+<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/69c2890">[69c2890]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script><script defer src=https://static.cloudflareinsights.com/beacon.min.js data-cf-beacon='{"token": "987638e636ce4dbb932d038af74c17d1"}'></script></body></html>
--- a/about/index.html
+++ b/about/index.html
@@ -0,0 +1,7 @@
+<!doctype html><html lang=en><head><title>About · Eric X. Liu's Personal Page</title><meta charset=utf-8><meta name=viewport content="width=device-width,initial-scale=1"><meta name=color-scheme content="light dark"><meta name=author content="Eric X. Liu"><meta name=description content="Eric X. Liu - Software & Performance Engineer at Google. Sharing insights about software engineering, performance optimization, tech industry experiences, mountain biking adventures, Jeep overlanding, and outdoor activities."><meta name=keywords content="software engineer,performance engineering,Google engineer,tech blog,software development,performance optimization,Eric Liu,engineering blog,mountain biking,Jeep enthusiast,overlanding,camping,outdoor adventures"><meta name=twitter:card content="summary"><meta name=twitter:title content="About"><meta name=twitter:description content="Eric X. Liu - Software & Performance Engineer at Google. Sharing insights about software engineering, performance optimization, tech industry experiences, mountain biking adventures, Jeep overlanding, and outdoor activities."><meta property="og:url" content="/about/"><meta property="og:site_name" content="Eric X. Liu's Personal Page"><meta property="og:title" content="About"><meta property="og:description" content="Eric X. Liu - Software & Performance Engineer at Google. Sharing insights about software engineering, performance optimization, tech industry experiences, mountain biking adventures, Jeep overlanding, and outdoor activities."><meta property="og:locale" content="en"><meta property="og:type" content="article"><meta property="article:published_time" content="2018-06-01T07:13:52+00:00"><meta property="article:modified_time" content="2020-06-16T23:30:17-07:00"><link rel=canonical href=/about/><link rel=preload href=/fonts/fa-brands-400.woff2 as=font type=font/woff2 crossorigin><link rel=preload href=/fonts/fa-regular-400.woff2 as=font type=font/woff2 crossorigin><link rel=preload href=/fonts/fa-solid-900.woff2 as=font type=font/woff2 crossorigin><link rel=stylesheet href=/css/coder.min.6445a802b9389c9660e1b07b724dcf5718b1065ed2d71b4eeaf981cc7cc5fc46.css integrity="sha256-ZEWoArk4nJZg4bB7ck3PVxixBl7S1xtO6vmBzHzF/EY=" crossorigin=anonymous media=screen><link rel=stylesheet href=/css/coder-dark.min.a00e6364bacbc8266ad1cc81230774a1397198f8cfb7bcba29b7d6fcb54ce57f.css integrity="sha256-oA5jZLrLyCZq0cyBIwd0oTlxmPjPt7y6KbfW/LVM5X8=" crossorigin=anonymous media=screen><link rel=icon type=image/svg+xml href=/images/favicon.svg sizes=any><link rel=icon type=image/png href=/images/favicon-32x32.png sizes=32x32><link rel=icon type=image/png href=/images/favicon-16x16.png sizes=16x16><link rel=apple-touch-icon href=/images/apple-touch-icon.png><link rel=apple-touch-icon sizes=180x180 href=/images/apple-touch-icon.png><link rel=manifest href=/site.webmanifest><link rel=mask-icon href=/images/safari-pinned-tab.svg color=#5bbad5></head><body class="preload-transitions colorscheme-auto"><div class=float-container><a id=dark-mode-toggle class=colorscheme-toggle><i class="fa-solid fa-adjust fa-fw" aria-hidden=true></i></a></div><main class=wrapper><nav class=navigation><section class=container><a class=navigation-title href=/>Eric X. Liu's Personal Page
+</a><input type=checkbox id=menu-toggle>
+<label class="menu-button float-right" for=menu-toggle><i class="fa-solid fa-bars fa-fw" aria-hidden=true></i></label><ul class=navigation-list><li class=navigation-item><a class=navigation-link href=/posts/>Posts</a></li><li class=navigation-item><a class=navigation-link href=https://chat.ericxliu.me>Chat</a></li><li class=navigation-item><a class=navigation-link href=https://git.ericxliu.me/user/oauth2/Authenitk>Git</a></li><li class=navigation-item><a class=navigation-link href=https://coder.ericxliu.me/api/v2/users/oidc/callback>Coder</a></li><li class=navigation-item><a class=navigation-link href=/>|</a></li><li class=navigation-item><a class=navigation-link href=https://sso.ericxliu.me>Sign in</a></li></ul></section></nav><div class=content><section class="container page"><article><header><h1 class=title><a class=title-link href=/about/>About</a></h1></header></article></section><link rel=stylesheet href=https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/katex.min.css integrity=sha384-vKruj+a13U8yHIkAyGgK1J3ArTLzrFGBbBc0tDp4ad/EyewESeXE/Iv67Aj8gKZ0 crossorigin=anonymous><script defer src=https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/katex.min.js integrity=sha384-PwRUT/YqbnEjkZO0zZxNqcxACrXe+j766U2amXcgMg5457rve2Y7I6ZJSm2A0mS4 crossorigin=anonymous></script><script defer src=https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/contrib/auto-render.min.js integrity=sha384-+VBxd3r6XgURycqtZ117nYw44OOcIax56Z4dCRWbxyPt0Koah1uHoK0o4+/RRE05 crossorigin=anonymous onload='renderMathInElement(document.body,{delimiters:[{left:"$$",right:"$$",display:!0},{left:"$",right:"$",display:!1},{left:"\\(",right:"\\)",display:!1},{left:"\\[",right:"\\]",display:!0}]})'></script></div><footer class=footer><section class=container>©
+2016 -
+2025
+Eric X. Liu
+<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/69c2890">[69c2890]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script><script defer src=https://static.cloudflareinsights.com/beacon.min.js data-cf-beacon='{"token": "987638e636ce4dbb932d038af74c17d1"}'></script></body></html>
--- a/archetypes/default.md
+++ b/archetypes/default.md
@@ -1,6 +0,0 @@
---
-title: "{{ replace .Name "-" " " | title }}"
-date: {{ .Date }}
-draft: true
---
-
--- a/categories/index.html
+++ b/categories/index.html
@@ -0,0 +1,7 @@
+<!doctype html><html lang=en><head><title>Categories · Eric X. Liu's Personal Page</title><meta charset=utf-8><meta name=viewport content="width=device-width,initial-scale=1"><meta name=color-scheme content="light dark"><meta name=author content="Eric X. Liu"><meta name=description content="Eric X. Liu - Software & Performance Engineer at Google. Sharing insights about software engineering, performance optimization, tech industry experiences, mountain biking adventures, Jeep overlanding, and outdoor activities."><meta name=keywords content="software engineer,performance engineering,Google engineer,tech blog,software development,performance optimization,Eric Liu,engineering blog,mountain biking,Jeep enthusiast,overlanding,camping,outdoor adventures"><meta name=twitter:card content="summary"><meta name=twitter:title content="Categories"><meta name=twitter:description content="Eric X. Liu - Software & Performance Engineer at Google. Sharing insights about software engineering, performance optimization, tech industry experiences, mountain biking adventures, Jeep overlanding, and outdoor activities."><meta property="og:url" content="/categories/"><meta property="og:site_name" content="Eric X. Liu's Personal Page"><meta property="og:title" content="Categories"><meta property="og:description" content="Eric X. Liu - Software & Performance Engineer at Google. Sharing insights about software engineering, performance optimization, tech industry experiences, mountain biking adventures, Jeep overlanding, and outdoor activities."><meta property="og:locale" content="en"><meta property="og:type" content="website"><link rel=canonical href=/categories/><link rel=preload href=/fonts/fa-brands-400.woff2 as=font type=font/woff2 crossorigin><link rel=preload href=/fonts/fa-regular-400.woff2 as=font type=font/woff2 crossorigin><link rel=preload href=/fonts/fa-solid-900.woff2 as=font type=font/woff2 crossorigin><link rel=stylesheet href=/css/coder.min.6445a802b9389c9660e1b07b724dcf5718b1065ed2d71b4eeaf981cc7cc5fc46.css integrity="sha256-ZEWoArk4nJZg4bB7ck3PVxixBl7S1xtO6vmBzHzF/EY=" crossorigin=anonymous media=screen><link rel=stylesheet href=/css/coder-dark.min.a00e6364bacbc8266ad1cc81230774a1397198f8cfb7bcba29b7d6fcb54ce57f.css integrity="sha256-oA5jZLrLyCZq0cyBIwd0oTlxmPjPt7y6KbfW/LVM5X8=" crossorigin=anonymous media=screen><link rel=icon type=image/svg+xml href=/images/favicon.svg sizes=any><link rel=icon type=image/png href=/images/favicon-32x32.png sizes=32x32><link rel=icon type=image/png href=/images/favicon-16x16.png sizes=16x16><link rel=apple-touch-icon href=/images/apple-touch-icon.png><link rel=apple-touch-icon sizes=180x180 href=/images/apple-touch-icon.png><link rel=manifest href=/site.webmanifest><link rel=mask-icon href=/images/safari-pinned-tab.svg color=#5bbad5><link rel=alternate type=application/rss+xml href=/categories/index.xml title="Eric X. Liu's Personal Page"></head><body class="preload-transitions colorscheme-auto"><div class=float-container><a id=dark-mode-toggle class=colorscheme-toggle><i class="fa-solid fa-adjust fa-fw" aria-hidden=true></i></a></div><main class=wrapper><nav class=navigation><section class=container><a class=navigation-title href=/>Eric X. Liu's Personal Page
+</a><input type=checkbox id=menu-toggle>
+<label class="menu-button float-right" for=menu-toggle><i class="fa-solid fa-bars fa-fw" aria-hidden=true></i></label><ul class=navigation-list><li class=navigation-item><a class=navigation-link href=/posts/>Posts</a></li><li class=navigation-item><a class=navigation-link href=https://chat.ericxliu.me>Chat</a></li><li class=navigation-item><a class=navigation-link href=https://git.ericxliu.me/user/oauth2/Authenitk>Git</a></li><li class=navigation-item><a class=navigation-link href=https://coder.ericxliu.me/api/v2/users/oidc/callback>Coder</a></li><li class=navigation-item><a class=navigation-link href=/>|</a></li><li class=navigation-item><a class=navigation-link href=https://sso.ericxliu.me>Sign in</a></li></ul></section></nav><div class=content><section class="container taxonomy"><header><h1 class=title><a class=title-link href=/categories/>Categories</a></h1></header><ul></ul></section></div><footer class=footer><section class=container>©
+2016 -
+2025
+Eric X. Liu
+<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/69c2890">[69c2890]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script><script defer src=https://static.cloudflareinsights.com/beacon.min.js data-cf-beacon='{"token": "987638e636ce4dbb932d038af74c17d1"}'></script></body></html>
--- a/categories/index.xml
+++ b/categories/index.xml
@@ -0,0 +1 @@
+<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Categories on Eric X. Liu's Personal Page</title><link>/categories/</link><description>Recent content in Categories on Eric X. Liu's Personal Page</description><generator>Hugo</generator><language>en</language><atom:link href="/categories/index.xml" rel="self" type="application/rss+xml"/></channel></rss>
--- a/config.toml
+++ b/config.toml
@@ -1,84 +0,0 @@
-title = "Eric's Personal Page"
-
-theme = "hugo-coder"
-
-languageCode = "en"
-defaultcontentlanguage = "en"
-
-paginate = 20
-canonifyurls = true
-
-pygmentsstyle = "b2"
-pygmentscodefences = true
-pygmentscodefencesguesssyntax = true
-
-[params] # theme parameters
-    author = "Eric Liu"
-    info = "Platform Software & Performance Engineer @Google"
-    description = "Eric Liu's personal website"
-    keywords = "blog,developer,personal" 
-    avatarurl = "images/gravatar.png"
-
-    # wether you want to hide copyright and credits in the footer
-    hideCredits = true
-    hideCopyright = false
-
-    rtl= false
-
-    colorscheme = "light"
-
-    # Series see also post count
-    maxSeeAlsoItems = 5
-
-    # Enable Twemoji
-    enableTwemoji = true
-
-    # Custom CSS
-    custom_css = []
-
-    # Custom JS
-    custom_js = []
-
-# Social links
-[[params.social]]
-    name = "Git"
-    icon = "fab fa-gitlab"
-    weight = 1
-    url = "https://git.ericxliu.me/eric"
-[[params.social]]
-    name = "linkedin"
-    icon = "fab fa-linkedin"
-    weight = 2
-    url = "https://www.linkedin.com/in/eric-liu-46648b93/"
-[[params.social]]
-    name = "Personal email"
-    icon = "fas fa-envelope-square"
-    weight = 3
-
-# Menu links
-[languages]
-    [languages.en]
-        languagename = "English"
-	[[languages.en.menu.main]]
-	    name = "Posts"
-	    weight = 1
-	    url  = "/posts/"
-	[[languages.en.menu.main]]
-	    name = "Gitlab"
-	    weight = 2
-	    url  = "https://git.ericxliu.me"
-	[[languages.en.menu.main]]
-	    name = "Notebook"
-	    weight = 3
-	    url  = "https://hub.ericxliu.me"
-	[[languages.en.menu.main]]
-	    name = "Go"
-	    weight = 4
-	    url  = "https://go.ericxliu.me/server"
-	[[languages.en.menu.main]]
-	    name = "|"
-	    weight = 10
-	[[languages.en.menu.main]]
-	    name = "Sign in"
-	    weight = 11
-	    url  = "https://auth.ericxliu.me"
--- a/content/about.md
+++ b/content/about.md
@@ -1,5 +0,0 @@
---
-title: "About"
-date: 2018-06-01T07:13:52Z
---
-
--- a/content/posts/credit_card.html
+++ b/content/posts/credit_card.html
--- a/content/posts/useful.md
+++ b/content/posts/useful.md
@@ -1,12 +0,0 @@
-+++ 
-date = 2020-10-26T04:14:43Z
-title = "Some useful files"
-description = ""
-slug = ""
-tags = []
-categories = []
-externalLink = ""
-series = []
-+++
-* [rootCA.pem](https://ericxliu.me/rootCA.pem)
-* [vpnclient.ovpn](https://ericxliu.me/vpnclient.ovpn)
--- a/css/coder-dark.min.a00e6364bacbc8266ad1cc81230774a1397198f8cfb7bcba29b7d6fcb54ce57f.css
+++ b/css/coder-dark.min.a00e6364bacbc8266ad1cc81230774a1397198f8cfb7bcba29b7d6fcb54ce57f.css
--- a/css/coder.min.6445a802b9389c9660e1b07b724dcf5718b1065ed2d71b4eeaf981cc7cc5fc46.css
+++ b/css/coder.min.6445a802b9389c9660e1b07b724dcf5718b1065ed2d71b4eeaf981cc7cc5fc46.css
--- a/fonts/LICENSE.txt
+++ b/fonts/LICENSE.txt
@@ -0,0 +1,165 @@
+Fonticons, Inc. (https://fontawesome.com)
+
+--------------------------------------------------------------------------------
+
+Font Awesome Free License
+
+Font Awesome Free is free, open source, and GPL friendly. You can use it for
+commercial projects, open source projects, or really almost whatever you want.
+Full Font Awesome Free license: https://fontawesome.com/license/free.
+
+--------------------------------------------------------------------------------
+
+# Icons: CC BY 4.0 License (https://creativecommons.org/licenses/by/4.0/)
+
+The Font Awesome Free download is licensed under a Creative Commons
+Attribution 4.0 International License and applies to all icons packaged
+as SVG and JS file types.
+
+--------------------------------------------------------------------------------
+
+# Fonts: SIL OFL 1.1 License
+
+In the Font Awesome Free download, the SIL OFL license applies to all icons
+packaged as web and desktop font files.
+
+Copyright (c) 2024 Fonticons, Inc. (https://fontawesome.com)
+with Reserved Font Name: "Font Awesome".
+
+This Font Software is licensed under the SIL Open Font License, Version 1.1.
+This license is copied below, and is also available with a FAQ at:
+http://scripts.sil.org/OFL
+
+SIL OPEN FONT LICENSE
+Version 1.1 - 26 February 2007
+
+PREAMBLE
+The goals of the Open Font License (OFL) are to stimulate worldwide
+development of collaborative font projects, to support the font creation
+efforts of academic and linguistic communities, and to provide a free and
+open framework in which fonts may be shared and improved in partnership
+with others.
+
+The OFL allows the licensed fonts to be used, studied, modified and
+redistributed freely as long as they are not sold by themselves. The
+fonts, including any derivative works, can be bundled, embedded,
+redistributed and/or sold with any software provided that any reserved
+names are not used by derivative works. The fonts and derivatives,
+however, cannot be released under any other type of license. The
+requirement for fonts to remain under this license does not apply
+to any document created using the fonts or their derivatives.
+
+DEFINITIONS
+"Font Software" refers to the set of files released by the Copyright
+Holder(s) under this license and clearly marked as such. This may
+include source files, build scripts and documentation.
+
+"Reserved Font Name" refers to any names specified as such after the
+copyright statement(s).
+
+"Original Version" refers to the collection of Font Software components as
+distributed by the Copyright Holder(s).
+
+"Modified Version" refers to any derivative made by adding to, deleting,
+or substituting — in part or in whole — any of the components of the
+Original Version, by changing formats or by porting the Font Software to a
+new environment.
+
+"Author" refers to any designer, engineer, programmer, technical
+writer or other person who contributed to the Font Software.
+
+PERMISSION & CONDITIONS
+Permission is hereby granted, free of charge, to any person obtaining
+a copy of the Font Software, to use, study, copy, merge, embed, modify,
+redistribute, and sell modified and unmodified copies of the Font
+Software, subject to the following conditions:
+
+1) Neither the Font Software nor any of its individual components,
+in Original or Modified Versions, may be sold by itself.
+
+2) Original or Modified Versions of the Font Software may be bundled,
+redistributed and/or sold with any software, provided that each copy
+contains the above copyright notice and this license. These can be
+included either as stand-alone text files, human-readable headers or
+in the appropriate machine-readable metadata fields within text or
+binary files as long as those fields can be easily viewed by the user.
+
+3) No Modified Version of the Font Software may use the Reserved Font
+Name(s) unless explicit written permission is granted by the corresponding
+Copyright Holder. This restriction only applies to the primary font name as
+presented to the users.
+
+4) The name(s) of the Copyright Holder(s) or the Author(s) of the Font
+Software shall not be used to promote, endorse or advertise any
+Modified Version, except to acknowledge the contribution(s) of the
+Copyright Holder(s) and the Author(s) or with their explicit written
+permission.
+
+5) The Font Software, modified or unmodified, in part or in whole,
+must be distributed entirely under this license, and must not be
+distributed under any other license. The requirement for fonts to
+remain under this license does not apply to any document created
+using the Font Software.
+
+TERMINATION
+This license becomes null and void if any of the above conditions are
+not met.
+
+DISCLAIMER
+THE FONT SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT
+OF COPYRIGHT, PATENT, TRADEMARK, OR OTHER RIGHT. IN NO EVENT SHALL THE
+COPYRIGHT HOLDER BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+INCLUDING ANY GENERAL, SPECIAL, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL
+DAMAGES, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+FROM, OUT OF THE USE OR INABILITY TO USE THE FONT SOFTWARE OR FROM
+OTHER DEALINGS IN THE FONT SOFTWARE.
+
+--------------------------------------------------------------------------------
+
+# Code: MIT License (https://opensource.org/licenses/MIT)
+
+In the Font Awesome Free download, the MIT license applies to all non-font and
+non-icon files.
+
+Copyright 2024 Fonticons, Inc.
+
+Permission is hereby granted, free of charge, to any person obtaining a copy of
+this software and associated documentation files (the "Software"), to deal in the
+Software without restriction, including without limitation the rights to use, copy,
+modify, merge, publish, distribute, sublicense, and/or sell copies of the Software,
+and to permit persons to whom the Software is furnished to do so, subject to the
+following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED,
+INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A
+PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
+HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+--------------------------------------------------------------------------------
+
+# Attribution
+
+Attribution is required by MIT, SIL OFL, and CC BY licenses. Downloaded Font
+Awesome Free files already contain embedded comments with sufficient
+attribution, so you shouldn't need to do anything additional when using these
+files normally.
+
+We've kept attribution comments terse, so we ask that you do not actively work
+to remove them from files, especially code. They're a great way for folks to
+learn about Font Awesome.
+
+--------------------------------------------------------------------------------
+
+# Brand Icons
+
+All brand icons are trademarks of their respective owners. The use of these
+trademarks does not indicate endorsement of the trademark holder by Font
+Awesome, nor vice versa. **Please do not use brand logos for any purpose except
+to represent the company, product, or service to which they refer.**
--- a/fonts/fa-brands-400.ttf
+++ b/fonts/fa-brands-400.ttf
--- a/fonts/fa-brands-400.woff2
+++ b/fonts/fa-brands-400.woff2
--- a/fonts/fa-regular-400.ttf
+++ b/fonts/fa-regular-400.ttf
--- a/fonts/fa-regular-400.woff2
+++ b/fonts/fa-regular-400.woff2
--- a/fonts/fa-solid-900.ttf
+++ b/fonts/fa-solid-900.ttf
--- a/fonts/fa-solid-900.woff2
+++ b/fonts/fa-solid-900.woff2
--- a/images/a-deep-dive-into-ppo-for-language-models/.png
+++ b/images/a-deep-dive-into-ppo-for-language-models/.png
--- a/images/a-deep-dive-into-ppo-for-language-models/64bfdb4b-678e-4bfc-8b62-0c05c243f6a9.png
+++ b/images/a-deep-dive-into-ppo-for-language-models/64bfdb4b-678e-4bfc-8b62-0c05c243f6a9.png
--- a/images/a-technical-deep-dive-into-the-transformer-s-core-mechanics/.png
+++ b/images/a-technical-deep-dive-into-the-transformer-s-core-mechanics/.png
--- a/images/gravatar.png
+++ b/images/gravatar.png
--- a/images/ppo-for-language-models/.png
+++ b/images/ppo-for-language-models/.png
--- a/images/transformer-s-core-mechanics/.png
+++ b/images/transformer-s-core-mechanics/.png
--- a/index.html
+++ b/index.html
@@ -0,0 +1,7 @@
+<!doctype html><html lang=en><head><title>Eric X. Liu's Personal Page</title><meta charset=utf-8><meta name=viewport content="width=device-width,initial-scale=1"><meta name=color-scheme content="light dark"><meta name=author content="Eric X. Liu"><meta name=description content="Eric X. Liu - Software & Performance Engineer at Google. Sharing insights about software engineering, performance optimization, tech industry experiences, mountain biking adventures, Jeep overlanding, and outdoor activities."><meta name=keywords content="software engineer,performance engineering,Google engineer,tech blog,software development,performance optimization,Eric Liu,engineering blog,mountain biking,Jeep enthusiast,overlanding,camping,outdoor adventures"><meta name=twitter:card content="summary"><meta name=twitter:title content="Eric X. Liu's Personal Page"><meta name=twitter:description content="Eric X. Liu - Software & Performance Engineer at Google. Sharing insights about software engineering, performance optimization, tech industry experiences, mountain biking adventures, Jeep overlanding, and outdoor activities."><meta property="og:url" content="/"><meta property="og:site_name" content="Eric X. Liu's Personal Page"><meta property="og:title" content="Eric X. Liu's Personal Page"><meta property="og:description" content="Eric X. Liu - Software & Performance Engineer at Google. Sharing insights about software engineering, performance optimization, tech industry experiences, mountain biking adventures, Jeep overlanding, and outdoor activities."><meta property="og:locale" content="en"><meta property="og:type" content="website"><link rel=canonical href=/><link rel=preload href=/fonts/fa-brands-400.woff2 as=font type=font/woff2 crossorigin><link rel=preload href=/fonts/fa-regular-400.woff2 as=font type=font/woff2 crossorigin><link rel=preload href=/fonts/fa-solid-900.woff2 as=font type=font/woff2 crossorigin><link rel=stylesheet href=/css/coder.min.6445a802b9389c9660e1b07b724dcf5718b1065ed2d71b4eeaf981cc7cc5fc46.css integrity="sha256-ZEWoArk4nJZg4bB7ck3PVxixBl7S1xtO6vmBzHzF/EY=" crossorigin=anonymous media=screen><link rel=stylesheet href=/css/coder-dark.min.a00e6364bacbc8266ad1cc81230774a1397198f8cfb7bcba29b7d6fcb54ce57f.css integrity="sha256-oA5jZLrLyCZq0cyBIwd0oTlxmPjPt7y6KbfW/LVM5X8=" crossorigin=anonymous media=screen><link rel=icon type=image/svg+xml href=/images/favicon.svg sizes=any><link rel=icon type=image/png href=/images/favicon-32x32.png sizes=32x32><link rel=icon type=image/png href=/images/favicon-16x16.png sizes=16x16><link rel=apple-touch-icon href=/images/apple-touch-icon.png><link rel=apple-touch-icon sizes=180x180 href=/images/apple-touch-icon.png><link rel=manifest href=/site.webmanifest><link rel=mask-icon href=/images/safari-pinned-tab.svg color=#5bbad5><link rel=alternate type=application/rss+xml href=/index.xml title="Eric X. Liu's Personal Page"><meta name=generator content="Hugo 0.148.2"></head><body class="preload-transitions colorscheme-auto"><div class=float-container><a id=dark-mode-toggle class=colorscheme-toggle><i class="fa-solid fa-adjust fa-fw" aria-hidden=true></i></a></div><main class=wrapper><nav class=navigation><section class=container><a class=navigation-title href=/>Eric X. Liu's Personal Page
+</a><input type=checkbox id=menu-toggle>
+<label class="menu-button float-right" for=menu-toggle><i class="fa-solid fa-bars fa-fw" aria-hidden=true></i></label><ul class=navigation-list><li class=navigation-item><a class=navigation-link href=/posts/>Posts</a></li><li class=navigation-item><a class=navigation-link href=https://chat.ericxliu.me>Chat</a></li><li class=navigation-item><a class=navigation-link href=https://git.ericxliu.me/user/oauth2/Authenitk>Git</a></li><li class=navigation-item><a class=navigation-link href=https://coder.ericxliu.me/api/v2/users/oidc/callback>Coder</a></li><li class=navigation-item><a class=navigation-link href=/>|</a></li><li class=navigation-item><a class=navigation-link href=https://sso.ericxliu.me>Sign in</a></li></ul></section></nav><div class=content><section class="container centered"><div class=about><div class=avatar><img src=/images/gravatar.png alt=avatar></div><h1>Eric X. Liu</h1><h2>Software & Performance Engineer @Google</h2><ul><li><a href=https://git.ericxliu.me/eric aria-label=Git><i class="fa-brands fa-git fa-2x" aria-hidden=true></i></a></li><li><a href=https://www.linkedin.com/in/eric-x-liu-46648b93/ aria-label=linkedin><i class="fa-brands fa-linkedin fa-2x" aria-hidden=true></i></a></li><li><a href aria-label="Personal email"><i class="fa fa-envelope fa-2x" aria-hidden=true></i></a></li></ul></div></section></div><footer class=footer><section class=container>©
+2016 -
+2025
+Eric X. Liu
+<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/69c2890">[69c2890]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script><script defer src=https://static.cloudflareinsights.com/beacon.min.js data-cf-beacon='{"token": "987638e636ce4dbb932d038af74c17d1"}'></script></body></html>
--- a/index.xml
+++ b/index.xml
@@ -0,0 +1,38 @@
+<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Eric X. Liu's Personal Page</title><link>/</link><description>Recent content on Eric X. Liu's Personal Page</description><generator>Hugo</generator><language>en</language><lastBuildDate>Wed, 20 Aug 2025 06:28:39 +0000</lastBuildDate><atom:link href="/index.xml" rel="self" type="application/rss+xml"/><item><title>Quantization in LLMs</title><link>/posts/quantization-in-llms/</link><pubDate>Tue, 19 Aug 2025 00:00:00 +0000</pubDate><guid>/posts/quantization-in-llms/</guid><description>&lt;p>The burgeoning scale of Large Language Models (LLMs) has necessitated a paradigm shift in their deployment, moving beyond full-precision floating-point arithmetic towards lower-precision representations. Quantization, the process of mapping a wide range of continuous values to a smaller, discrete set, has emerged as a critical technique to reduce model size, accelerate inference, and lower energy consumption. This article provides a technical overview of quantization theories, their application in modern LLMs, and highlights the ongoing innovations in this domain.&lt;/p></description></item><item><title>Breville Barista Pro Maintenance</title><link>/posts/breville-barista-pro-maintenance/</link><pubDate>Sat, 16 Aug 2025 00:00:00 +0000</pubDate><guid>/posts/breville-barista-pro-maintenance/</guid><description>&lt;p>Proper maintenance is critical for the longevity and performance of a Breville Barista Pro espresso machine. Consistent cleaning not only ensures the machine functions correctly but also directly impacts the quality of the espresso produced. This guide provides a detailed, technical breakdown of the essential maintenance routines, from automated cycles to daily upkeep.&lt;/p>
+&lt;h4 id="understanding-the-two-primary-maintenance-cycles">
+ &lt;strong>Understanding the Two Primary Maintenance Cycles&lt;/strong>
+ &lt;a class="heading-link" href="#understanding-the-two-primary-maintenance-cycles">
+ &lt;i class="fa-solid fa-link" aria-hidden="true" title="Link to heading">&lt;/i>
+ &lt;span class="sr-only">Link to heading&lt;/span>
+ &lt;/a>
+&lt;/h4>
+&lt;p>The Breville Barista Pro has two distinct, automated maintenance procedures: the &lt;strong>Cleaning (Flush) Cycle&lt;/strong> and the &lt;strong>Descale Cycle&lt;/strong>. It is important to understand that these are not interchangeable, as they address different types of buildup within the machine.&lt;/p></description></item><item><title>Fixing GPU Operator Pods Stuck in Init: Secure Boot, DKMS, and MOK on Proxmox + Debian</title><link>/posts/secure-boot-dkms-and-mok-on-proxmox-debian/</link><pubDate>Sat, 09 Aug 2025 00:00:00 +0000</pubDate><guid>/posts/secure-boot-dkms-and-mok-on-proxmox-debian/</guid><description>&lt;p>I hit an issue where all GPU Operator pods on one node were stuck in Init after migrating from Legacy BIOS to UEFI. The common error was NVIDIA components waiting for “toolkit-ready,” while the toolkit init container looped with:&lt;/p>
+&lt;ul>
+&lt;li>nvidia-smi failed to communicate with the NVIDIA driver&lt;/li>
+&lt;li>modprobe nvidia → “Key was rejected by service”&lt;/li>
+&lt;/ul>
+&lt;p>That message is the tell: Secure Boot is enabled and the kernel refuses to load modules not signed by a trusted key.&lt;/p></description></item><item><title>Beyond Words: How RVQ Teaches LLMs to See and Hear</title><link>/posts/how-rvq-teaches-llms-to-see-and-hear/</link><pubDate>Thu, 07 Aug 2025 00:00:00 +0000</pubDate><guid>/posts/how-rvq-teaches-llms-to-see-and-hear/</guid><description>&lt;p>Large Language Models (LLMs) are masters of text, but the world is not made of text alone. It’s a symphony of sights, sounds, and experiences. The ultimate goal for AI is to understand this rich, multi-modal world as we do. But how do you teach a model that thinks in words to understand a picture of a sunset or the melody of a song?&lt;/p>
+&lt;p>The answer lies in creating a universal language—a bridge between the continuous, messy world of pixels and audio waves and the discrete, structured world of language tokens. One of the most elegant and powerful tools for building this bridge is &lt;strong>Residual Vector Quantization (RVQ)&lt;/strong>.&lt;/p></description></item><item><title>Supabase Deep Dive: It's Not Magic, It's Just Postgres</title><link>/posts/supabase-deep-dive/</link><pubDate>Sun, 03 Aug 2025 00:00:00 +0000</pubDate><guid>/posts/supabase-deep-dive/</guid><description>&lt;p>In the world of Backend-as-a-Service (BaaS), platforms are often treated as magic boxes. You push data in, you get data out, and you hope the magic inside scales. While this simplicity is powerful, it can obscure the underlying mechanics, leaving developers wondering what&amp;rsquo;s really going on.&lt;/p>
+&lt;p>Supabase enters this space with a radically different philosophy: &lt;strong>transparency&lt;/strong>. It provides the convenience of a BaaS, but it’s built on the world&amp;rsquo;s most trusted relational database: PostgreSQL. The &amp;ldquo;magic&amp;rdquo; isn&amp;rsquo;t a proprietary black box; it&amp;rsquo;s a carefully assembled suite of open-source tools that enhance Postgres, not hide it.&lt;/p></description></item><item><title>A Deep Dive into PPO for Language Models</title><link>/posts/ppo-for-language-models/</link><pubDate>Sat, 02 Aug 2025 00:00:00 +0000</pubDate><guid>/posts/ppo-for-language-models/</guid><description>&lt;p>Large Language Models (LLMs) have demonstrated astonishing capabilities, but out-of-the-box, they are simply powerful text predictors. They don&amp;rsquo;t inherently understand what makes a response helpful, harmless, or aligned with human values. The technique that has proven most effective at bridging this gap is Reinforcement Learning from Human Feedback (RLHF), and at its heart lies a powerful algorithm: Proximal Policy Optimization (PPO).&lt;/p>
+&lt;p>You may have seen diagrams like the one below, which outlines the RLHF training process. It can look intimidating, with a web of interconnected models, losses, and data flows.&lt;/p></description></item><item><title>Mixture-of-Experts (MoE) Models Challenges &amp; Solutions in Practice</title><link>/posts/mixture-of-experts-moe-models-challenges-solutions-in-practice/</link><pubDate>Wed, 02 Jul 2025 00:00:00 +0000</pubDate><guid>/posts/mixture-of-experts-moe-models-challenges-solutions-in-practice/</guid><description>&lt;p>Mixture-of-Experts (MoEs) are neural network architectures that allow different parts of the model (called &amp;ldquo;experts&amp;rdquo;) to specialize in different types of inputs. A &amp;ldquo;gating network&amp;rdquo; or &amp;ldquo;router&amp;rdquo; learns to dispatch each input (or &amp;ldquo;token&amp;rdquo;) to a subset of these experts. While powerful for scaling models, MoEs introduce several practical challenges.&lt;/p>
+&lt;h3 id="1-challenge-non-differentiability-of-routing-functions">
+ 1. Challenge: Non-Differentiability of Routing Functions
+ &lt;a class="heading-link" href="#1-challenge-non-differentiability-of-routing-functions">
+ &lt;i class="fa-solid fa-link" aria-hidden="true" title="Link to heading">&lt;/i>
+ &lt;span class="sr-only">Link to heading&lt;/span>
+ &lt;/a>
+&lt;/h3>
+&lt;p>&lt;strong>The Problem:&lt;/strong>
+Many routing mechanisms, especially &amp;ldquo;Top-K routing,&amp;rdquo; involve a discrete, hard selection process. A common function is &lt;code>KeepTopK(v, k)&lt;/code>, which selects the top &lt;code>k&lt;/code> scoring elements from a vector &lt;code>v&lt;/code> and sets others to $-\infty$ or $0$.&lt;/p></description></item><item><title>An Architectural Deep Dive of T5</title><link>/posts/t5-the-transformer-that-zigged-when-others-zagged-an-architectural-deep-dive/</link><pubDate>Sun, 01 Jun 2025 00:00:00 +0000</pubDate><guid>/posts/t5-the-transformer-that-zigged-when-others-zagged-an-architectural-deep-dive/</guid><description>&lt;p>In the rapidly evolving landscape of Large Language Models, a few key architectures define the dominant paradigms. Today, the &amp;ldquo;decoder-only&amp;rdquo; model, popularized by the GPT series and its successors like LLaMA and Mistral, reigns supreme. These models are scaled to incredible sizes and excel at in-context learning.&lt;/p>
+&lt;p>But to truly understand the field, we must look at the pivotal models that explored different paths. Google&amp;rsquo;s T5, or &lt;strong>Text-to-Text Transfer Transformer&lt;/strong>, stands out as one of the most influential. It didn&amp;rsquo;t just introduce a new model; it proposed a new philosophy. This article dives deep into the architecture of T5, how it fundamentally differs from modern LLMs, and the lasting legacy of its unique design choices.&lt;/p></description></item><item><title>Mastering Your Breville Barista Pro: The Ultimate Guide to Dialing In Espresso</title><link>/posts/espresso-theory-application-a-guide-for-the-breville-barista-pro/</link><pubDate>Thu, 01 May 2025 00:00:00 +0000</pubDate><guid>/posts/espresso-theory-application-a-guide-for-the-breville-barista-pro/</guid><description>&lt;p>Are you ready to transform your home espresso game from good to genuinely great? The Breville Barista Pro is a fantastic machine, but unlocking its full potential requires understanding a few key principles. This guide will walk you through the systematic process of dialing in your espresso, ensuring every shot is delicious and repeatable.&lt;/p>
+&lt;p>Our overarching philosophy is simple: &lt;strong>isolate and change only one variable at a time.&lt;/strong> While numbers are crucial, your palate is the ultimate judge. Dose, ratio, and time are interconnected, but your &lt;strong>grind size&lt;/strong> is your most powerful lever.&lt;/p></description></item><item><title>Transformer's Core Mechanics</title><link>/posts/transformer-s-core-mechanics/</link><pubDate>Tue, 01 Apr 2025 00:00:00 +0000</pubDate><guid>/posts/transformer-s-core-mechanics/</guid><description>&lt;p>The Transformer architecture is the bedrock of modern Large Language Models (LLMs). While its high-level success is widely known, a deeper understanding requires dissecting its core components. This article provides a detailed, technical breakdown of the fundamental concepts within a Transformer block, from the notion of &amp;ldquo;channels&amp;rdquo; to the intricate workings of the attention mechanism and its relationship with other advanced architectures like Mixture of Experts.&lt;/p>
+&lt;h3 id="1-the-channel-a-foundational-view-of-d_model">
+ 1. The &amp;ldquo;Channel&amp;rdquo;: A Foundational View of &lt;code>d_model&lt;/code>
+ &lt;a class="heading-link" href="#1-the-channel-a-foundational-view-of-d_model">
+ &lt;i class="fa-solid fa-link" aria-hidden="true" title="Link to heading">&lt;/i>
+ &lt;span class="sr-only">Link to heading&lt;/span>
+ &lt;/a>
+&lt;/h3>
+&lt;p>In deep learning, a &amp;ldquo;channel&amp;rdquo; can be thought of as a feature dimension. While this term is common in Convolutional Neural Networks for images (e.g., Red, Green, Blue channels), in LLMs, the analogous concept is the model&amp;rsquo;s primary embedding dimension, commonly referred to as &lt;code>d_model&lt;/code>.&lt;/p></description></item><item><title>Some useful files</title><link>/posts/useful/</link><pubDate>Mon, 26 Oct 2020 04:14:43 +0000</pubDate><guid>/posts/useful/</guid><description>&lt;ul>
+&lt;li>&lt;a href="/rootCA.crt" >rootCA.pem&lt;/a>&lt;/li>
+&lt;/ul></description></item><item><title>About</title><link>/about/</link><pubDate>Fri, 01 Jun 2018 07:13:52 +0000</pubDate><guid>/about/</guid><description/></item></channel></rss>
--- a/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js
+++ b/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js
@@ -0,0 +1 @@
+const body=document.body,darkModeToggle=document.getElementById("dark-mode-toggle"),darkModeMediaQuery=window.matchMedia("(prefers-color-scheme: dark)");localStorage.getItem("colorscheme")?setTheme(localStorage.getItem("colorscheme")):setTheme(body.classList.contains("colorscheme-light")||body.classList.contains("colorscheme-dark")?body.classList.contains("colorscheme-dark")?"dark":"light":darkModeMediaQuery.matches?"dark":"light"),darkModeToggle&&darkModeToggle.addEventListener("click",()=>{let e=body.classList.contains("colorscheme-dark")?"light":"dark";setTheme(e),rememberTheme(e)}),darkModeMediaQuery.addListener(e=>{setTheme(e.matches?"dark":"light")}),document.addEventListener("DOMContentLoaded",function(){let e=document.querySelector(".preload-transitions");e.classList.remove("preload-transitions")});function setTheme(e){body.classList.remove("colorscheme-auto");let n=e==="dark"?"light":"dark";body.classList.remove("colorscheme-"+n),body.classList.add("colorscheme-"+e),document.documentElement.style["color-scheme"]=e;function t(e){return new Promise(t=>{if(document.querySelector(e))return t(document.querySelector(e));const n=new MutationObserver(s=>{document.querySelector(e)&&(t(document.querySelector(e)),n.disconnect())});n.observe(document.body,{childList:!0,subtree:!0})})}if(e==="dark"){const e={type:"set-theme",theme:"github-dark"};t(".utterances-frame").then(t=>{t.contentWindow.postMessage(e,"https://utteranc.es")})}else{const e={type:"set-theme",theme:"github-light"};t(".utterances-frame").then(t=>{t.contentWindow.postMessage(e,"https://utteranc.es")})}function s(e){const t=document.querySelector("iframe.giscus-frame");if(!t)return;t.contentWindow.postMessage({giscus:e},"https://giscus.app")}s({setConfig:{theme:e}});const o=new Event("themeChanged");document.dispatchEvent(o)}function rememberTheme(e){localStorage.setItem("colorscheme",e)}
--- a/posts/breville-barista-pro-maintenance/index.html
+++ b/posts/breville-barista-pro-maintenance/index.html
@@ -0,0 +1,28 @@
+<!doctype html><html lang=en><head><title>Breville Barista Pro Maintenance · Eric X. Liu's Personal Page</title><meta charset=utf-8><meta name=viewport content="width=device-width,initial-scale=1"><meta name=color-scheme content="light dark"><meta name=author content="Eric X. Liu"><meta name=description content="Proper maintenance is critical for the longevity and performance of a Breville Barista Pro espresso machine. Consistent cleaning not only ensures the machine functions correctly but also directly impacts the quality of the espresso produced. This guide provides a detailed, technical breakdown of the essential maintenance routines, from automated cycles to daily upkeep.
+
+  Understanding the Two Primary Maintenance Cycles
+  
+    
+    Link to heading
+  
+
+The Breville Barista Pro has two distinct, automated maintenance procedures: the Cleaning (Flush) Cycle and the Descale Cycle. It is important to understand that these are not interchangeable, as they address different types of buildup within the machine."><meta name=keywords content="software engineer,performance engineering,Google engineer,tech blog,software development,performance optimization,Eric Liu,engineering blog,mountain biking,Jeep enthusiast,overlanding,camping,outdoor adventures"><meta name=twitter:card content="summary"><meta name=twitter:title content="Breville Barista Pro Maintenance"><meta name=twitter:description content="Proper maintenance is critical for the longevity and performance of a Breville Barista Pro espresso machine. Consistent cleaning not only ensures the machine functions correctly but also directly impacts the quality of the espresso produced. This guide provides a detailed, technical breakdown of the essential maintenance routines, from automated cycles to daily upkeep.
+Understanding the Two Primary Maintenance Cycles Link to heading The Breville Barista Pro has two distinct, automated maintenance procedures: the Cleaning (Flush) Cycle and the Descale Cycle. It is important to understand that these are not interchangeable, as they address different types of buildup within the machine."><meta property="og:url" content="/posts/breville-barista-pro-maintenance/"><meta property="og:site_name" content="Eric X. Liu's Personal Page"><meta property="og:title" content="Breville Barista Pro Maintenance"><meta property="og:description" content="Proper maintenance is critical for the longevity and performance of a Breville Barista Pro espresso machine. Consistent cleaning not only ensures the machine functions correctly but also directly impacts the quality of the espresso produced. This guide provides a detailed, technical breakdown of the essential maintenance routines, from automated cycles to daily upkeep.
+Understanding the Two Primary Maintenance Cycles Link to heading The Breville Barista Pro has two distinct, automated maintenance procedures: the Cleaning (Flush) Cycle and the Descale Cycle. It is important to understand that these are not interchangeable, as they address different types of buildup within the machine."><meta property="og:locale" content="en"><meta property="og:type" content="article"><meta property="article:section" content="posts"><meta property="article:published_time" content="2025-08-16T00:00:00+00:00"><meta property="article:modified_time" content="2025-08-20T06:04:36+00:00"><link rel=canonical href=/posts/breville-barista-pro-maintenance/><link rel=preload href=/fonts/fa-brands-400.woff2 as=font type=font/woff2 crossorigin><link rel=preload href=/fonts/fa-regular-400.woff2 as=font type=font/woff2 crossorigin><link rel=preload href=/fonts/fa-solid-900.woff2 as=font type=font/woff2 crossorigin><link rel=stylesheet href=/css/coder.min.6445a802b9389c9660e1b07b724dcf5718b1065ed2d71b4eeaf981cc7cc5fc46.css integrity="sha256-ZEWoArk4nJZg4bB7ck3PVxixBl7S1xtO6vmBzHzF/EY=" crossorigin=anonymous media=screen><link rel=stylesheet href=/css/coder-dark.min.a00e6364bacbc8266ad1cc81230774a1397198f8cfb7bcba29b7d6fcb54ce57f.css integrity="sha256-oA5jZLrLyCZq0cyBIwd0oTlxmPjPt7y6KbfW/LVM5X8=" crossorigin=anonymous media=screen><link rel=icon type=image/svg+xml href=/images/favicon.svg sizes=any><link rel=icon type=image/png href=/images/favicon-32x32.png sizes=32x32><link rel=icon type=image/png href=/images/favicon-16x16.png sizes=16x16><link rel=apple-touch-icon href=/images/apple-touch-icon.png><link rel=apple-touch-icon sizes=180x180 href=/images/apple-touch-icon.png><link rel=manifest href=/site.webmanifest><link rel=mask-icon href=/images/safari-pinned-tab.svg color=#5bbad5></head><body class="preload-transitions colorscheme-auto"><div class=float-container><a id=dark-mode-toggle class=colorscheme-toggle><i class="fa-solid fa-adjust fa-fw" aria-hidden=true></i></a></div><main class=wrapper><nav class=navigation><section class=container><a class=navigation-title href=/>Eric X. Liu's Personal Page
+</a><input type=checkbox id=menu-toggle>
+<label class="menu-button float-right" for=menu-toggle><i class="fa-solid fa-bars fa-fw" aria-hidden=true></i></label><ul class=navigation-list><li class=navigation-item><a class=navigation-link href=/posts/>Posts</a></li><li class=navigation-item><a class=navigation-link href=https://chat.ericxliu.me>Chat</a></li><li class=navigation-item><a class=navigation-link href=https://git.ericxliu.me/user/oauth2/Authenitk>Git</a></li><li class=navigation-item><a class=navigation-link href=https://coder.ericxliu.me/api/v2/users/oidc/callback>Coder</a></li><li class=navigation-item><a class=navigation-link href=/>|</a></li><li class=navigation-item><a class=navigation-link href=https://sso.ericxliu.me>Sign in</a></li></ul></section></nav><div class=content><section class="container post"><article><header><div class=post-title><h1 class=title><a class=title-link href=/posts/breville-barista-pro-maintenance/>Breville Barista Pro Maintenance</a></h1></div><div class=post-meta><div class=date><span class=posted-on><i class="fa-solid fa-calendar" aria-hidden=true></i>
+<time datetime=2025-08-16T00:00:00Z>August 16, 2025
+</time></span><span class=reading-time><i class="fa-solid fa-clock" aria-hidden=true></i>
+5-minute read</span></div></div></header><div class=post-content><p>Proper maintenance is critical for the longevity and performance of a Breville Barista Pro espresso machine. Consistent cleaning not only ensures the machine functions correctly but also directly impacts the quality of the espresso produced. This guide provides a detailed, technical breakdown of the essential maintenance routines, from automated cycles to daily upkeep.</p><h4 id=understanding-the-two-primary-maintenance-cycles><strong>Understanding the Two Primary Maintenance Cycles</strong>
+<a class=heading-link href=#understanding-the-two-primary-maintenance-cycles><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h4><p>The Breville Barista Pro has two distinct, automated maintenance procedures: the <strong>Cleaning (Flush) Cycle</strong> and the <strong>Descale Cycle</strong>. It is important to understand that these are not interchangeable, as they address different types of buildup within the machine.</p><ul><li><strong>Cleaning Cycle (Flush):</strong> This process is designed to remove coffee oils and granulated residue from the group head, shower screen, and portafilter system.</li><li><strong>Descale Cycle:</strong> This process targets the internal components of the machine, such as the thermocoil and water lines, to remove mineral and limescale deposits from water.</li></ul><h4 id=procedure-1-the-cleaning-flush-cycle><strong>Procedure 1: The Cleaning (Flush) Cycle</strong>
+<a class=heading-link href=#procedure-1-the-cleaning-flush-cycle><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h4><p>The machine will indicate when a cleaning cycle is needed by displaying a &ldquo;FLUSH&rdquo; alert on the LCD screen. This typically occurs after approximately 200 extractions.</p><p><strong>Required Materials:</strong></p><ul><li>1-Cup filter basket</li><li>Grey silicone cleaning disc (provided with the machine)</li><li>One cleaning tablet</li></ul><p><strong>Step-by-Step Instructions:</strong></p><ol><li>Insert the 1-cup filter basket into the portafilter.</li><li>Place the grey silicone cleaning disc inside the basket.</li><li>Position one cleaning tablet in the center of the disc.</li><li>Lock the portafilter firmly into the group head.</li><li>Ensure the drip tray is empty and the water tank is filled.</li><li>Press the &lsquo;MENU&rsquo; button and use the &lsquo;Grind Amount&rsquo; dial to navigate to the &lsquo;FLUSH&rsquo; option. Press the dial to select it.</li><li>The &lsquo;1 CUP&rsquo; button will illuminate. Press it to initiate the cycle.</li><li>The cleaning process will last approximately five minutes, with the machine backflushing water under pressure. The remaining time will be displayed on the screen.</li><li>Upon completion, the machine will beep and return to its ready state.</li><li>Remove the portafilter and discard the water and dissolved tablet residue. Thoroughly rinse the portafilter, cleaning disc, and filter basket.</li><li>Re-insert the portafilter (without the disc or tablet) and run a shot of hot water through the group head to rinse any remaining cleaning solution.</li></ol><h4 id=procedure-2-the-descale-cycle><strong>Procedure 2: The Descale Cycle</strong>
+<a class=heading-link href=#procedure-2-the-descale-cycle><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h4><p>The machine will alert you when descaling is required. The frequency depends on water hardness and usage but is generally recommended every 2-3 months.</p><p><strong>Required Materials:</strong></p><ul><li>Breville-recommended descaling solution</li><li>A large container (minimum 2-liter capacity)</li></ul><p><strong>Step-by-Step Instructions:</strong></p><p><strong>Part A: Preparation</strong></p><ol><li>Empty the drip tray and re-insert it.</li><li>Remove the water filter from the water tank.</li><li>Pour the descaling solution into the empty water tank and add fresh water up to the indicated &ldquo;DESCALE&rdquo; line.</li><li>Place a large container under the group head, hot water outlet, and steam wand.</li></ol><p><strong>Part B: The Descaling Process</strong></p><ol><li>Turn the machine on and press the &lsquo;MENU&rsquo; button. Navigate to the &lsquo;DESCALE&rsquo; option and select it by pressing the dial.</li><li>Press the illuminated &lsquo;1 CUP&rsquo; button to begin.</li><li>The cycle proceeds in three stages. You must manually advance through them using the steam dial based on the LCD prompts:<ul><li><strong>Group Head (d3):</strong> The machine descales the coffee brewing components.</li><li><strong>Hot Water (d2):</strong> After a beep, the LCD shows &ldquo;d2&rdquo;. Turn the steam dial to the hot water position.</li><li><strong>Steam (d1):</strong> After another beep, the display reads &ldquo;d1&rdquo;. Turn the dial to the steam position.</li></ul></li></ol><p><strong>Part C: The Rinse Cycle</strong></p><ol><li>Once the descaling solution is expended, the machine will beep and prompt for a rinse cycle (&ldquo;r&rdquo;).</li><li>Empty the large container and rinse the water tank thoroughly.</li><li>Fill the water tank with fresh, cold water to the MAX line and re-insert it.</li><li>Place the empty container back under the outlets and press the &lsquo;1 CUP&rsquo; button.</li><li>The rinse cycle will mirror the descaling process, prompting you to engage the group head (&ldquo;r3&rdquo;), hot water (&ldquo;r2&rdquo;), and steam wand (&ldquo;r1&rdquo;) in sequence.</li><li>After the rinse is complete, the machine will exit the maintenance mode and return to its ready state.</li></ol><h4 id=routine-and-preventative-maintenance-schedule><strong>Routine and Preventative Maintenance Schedule</strong>
+<a class=heading-link href=#routine-and-preventative-maintenance-schedule><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h4><p>In addition to the automated cycles, regular manual cleaning is essential for machine health.</p><p><strong>Daily Tasks:</strong></p><ul><li><strong>Purge Group Head:</strong> After the final use of the day, run hot water through the group head (without the portafilter) to clear grounds.</li><li><strong>Clean Portafilter & Baskets:</strong> Do not let used coffee grounds sit in the portafilter. Rinse with hot water after every use.</li><li><strong>Clean Steam Wand:</strong> Immediately after texturing milk, wipe the wand with a damp cloth and purge steam for 2-3 seconds to clear internal passages.</li><li><strong>Empty Drip Tray:</strong> Empty and rinse the drip tray regularly.</li></ul><p><strong>Weekly Tasks:</strong></p><ul><li><strong>Soak Components:</strong> Remove the filter basket from the portafilter. Soak both components in a solution of hot water and a cleaning tablet (or specific espresso cleaner) for 20-30 minutes to dissolve accumulated coffee oils. Rinse thoroughly.</li><li><strong>Clean Grinder:</strong> Empty the bean hopper. Run the grinder to clear any remaining beans, then use a brush and/or vacuum to clean out fines and oil residue from the burrs and chute.</li></ul><p><strong>Periodic Tasks (Every 2-3 Months):</strong></p><ul><li><strong>Replace Water Filter:</strong> The water filter located inside the water tank should be replaced every 3 months. This reduces the rate of scale buildup.</li><li><strong>Inspect Shower Screen:</strong> Use a brush to gently scrub the shower screen inside the group head to remove any stubborn coffee grounds.</li></ul><p>By adhering to this comprehensive maintenance schedule, you can ensure your Breville Barista Pro operates at peak performance and consistently produces high-quality espresso.</p><hr><p><strong>Reference:</strong></p><ul><li>Breville Barista Pro Instruction Manual and official manufacturer guidelines.</li></ul></div><footer><div id=disqus_thread></div><script>window.disqus_config=function(){},function(){if(["localhost","127.0.0.1"].indexOf(window.location.hostname)!=-1){document.getElementById("disqus_thread").innerHTML="Disqus comments not available by default when the website is previewed locally.";return}var t=document,e=t.createElement("script");e.async=!0,e.src="//ericxliu-me.disqus.com/embed.js",e.setAttribute("data-timestamp",+new Date),(t.head||t.body).appendChild(e)}(),document.addEventListener("themeChanged",function(){document.readyState=="complete"&&DISQUS.reset({reload:!0,config:disqus_config})})</script></footer></article><link rel=stylesheet href=https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/katex.min.css integrity=sha384-vKruj+a13U8yHIkAyGgK1J3ArTLzrFGBbBc0tDp4ad/EyewESeXE/Iv67Aj8gKZ0 crossorigin=anonymous><script defer src=https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/katex.min.js integrity=sha384-PwRUT/YqbnEjkZO0zZxNqcxACrXe+j766U2amXcgMg5457rve2Y7I6ZJSm2A0mS4 crossorigin=anonymous></script><script defer src=https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/contrib/auto-render.min.js integrity=sha384-+VBxd3r6XgURycqtZ117nYw44OOcIax56Z4dCRWbxyPt0Koah1uHoK0o4+/RRE05 crossorigin=anonymous onload='renderMathInElement(document.body,{delimiters:[{left:"$$",right:"$$",display:!0},{left:"$",right:"$",display:!1},{left:"\\(",right:"\\)",display:!1},{left:"\\[",right:"\\]",display:!0}]})'></script></section></div><footer class=footer><section class=container>©
+2016 -
+2025
+Eric X. Liu
+<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/69c2890">[69c2890]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script><script defer src=https://static.cloudflareinsights.com/beacon.min.js data-cf-beacon='{"token": "987638e636ce4dbb932d038af74c17d1"}'></script></body></html>
--- a/posts/espresso-theory-application-a-guide-for-the-breville-barista-pro/index.html
+++ b/posts/espresso-theory-application-a-guide-for-the-breville-barista-pro/index.html
@@ -0,0 +1,23 @@
+<!doctype html><html lang=en><head><title>Mastering Your Breville Barista Pro: The Ultimate Guide to Dialing In Espresso · Eric X. Liu's Personal Page</title><meta charset=utf-8><meta name=viewport content="width=device-width,initial-scale=1"><meta name=color-scheme content="light dark"><meta name=author content="Eric X. Liu"><meta name=description content="Are you ready to transform your home espresso game from good to genuinely great? The Breville Barista Pro is a fantastic machine, but unlocking its full potential requires understanding a few key principles. This guide will walk you through the systematic process of dialing in your espresso, ensuring every shot is delicious and repeatable.
+Our overarching philosophy is simple: isolate and change only one variable at a time. While numbers are crucial, your palate is the ultimate judge. Dose, ratio, and time are interconnected, but your grind size is your most powerful lever."><meta name=keywords content="software engineer,performance engineering,Google engineer,tech blog,software development,performance optimization,Eric Liu,engineering blog,mountain biking,Jeep enthusiast,overlanding,camping,outdoor adventures"><meta name=twitter:card content="summary"><meta name=twitter:title content="Mastering Your Breville Barista Pro: The Ultimate Guide to Dialing In Espresso"><meta name=twitter:description content="Are you ready to transform your home espresso game from good to genuinely great? The Breville Barista Pro is a fantastic machine, but unlocking its full potential requires understanding a few key principles. This guide will walk you through the systematic process of dialing in your espresso, ensuring every shot is delicious and repeatable.
+Our overarching philosophy is simple: isolate and change only one variable at a time. While numbers are crucial, your palate is the ultimate judge. Dose, ratio, and time are interconnected, but your grind size is your most powerful lever."><meta property="og:url" content="/posts/espresso-theory-application-a-guide-for-the-breville-barista-pro/"><meta property="og:site_name" content="Eric X. Liu's Personal Page"><meta property="og:title" content="Mastering Your Breville Barista Pro: The Ultimate Guide to Dialing In Espresso"><meta property="og:description" content="Are you ready to transform your home espresso game from good to genuinely great? The Breville Barista Pro is a fantastic machine, but unlocking its full potential requires understanding a few key principles. This guide will walk you through the systematic process of dialing in your espresso, ensuring every shot is delicious and repeatable.
+Our overarching philosophy is simple: isolate and change only one variable at a time. While numbers are crucial, your palate is the ultimate judge. Dose, ratio, and time are interconnected, but your grind size is your most powerful lever."><meta property="og:locale" content="en"><meta property="og:type" content="article"><meta property="article:section" content="posts"><meta property="article:published_time" content="2025-05-01T00:00:00+00:00"><meta property="article:modified_time" content="2025-08-03T04:20:20+00:00"><link rel=canonical href=/posts/espresso-theory-application-a-guide-for-the-breville-barista-pro/><link rel=preload href=/fonts/fa-brands-400.woff2 as=font type=font/woff2 crossorigin><link rel=preload href=/fonts/fa-regular-400.woff2 as=font type=font/woff2 crossorigin><link rel=preload href=/fonts/fa-solid-900.woff2 as=font type=font/woff2 crossorigin><link rel=stylesheet href=/css/coder.min.6445a802b9389c9660e1b07b724dcf5718b1065ed2d71b4eeaf981cc7cc5fc46.css integrity="sha256-ZEWoArk4nJZg4bB7ck3PVxixBl7S1xtO6vmBzHzF/EY=" crossorigin=anonymous media=screen><link rel=stylesheet href=/css/coder-dark.min.a00e6364bacbc8266ad1cc81230774a1397198f8cfb7bcba29b7d6fcb54ce57f.css integrity="sha256-oA5jZLrLyCZq0cyBIwd0oTlxmPjPt7y6KbfW/LVM5X8=" crossorigin=anonymous media=screen><link rel=icon type=image/svg+xml href=/images/favicon.svg sizes=any><link rel=icon type=image/png href=/images/favicon-32x32.png sizes=32x32><link rel=icon type=image/png href=/images/favicon-16x16.png sizes=16x16><link rel=apple-touch-icon href=/images/apple-touch-icon.png><link rel=apple-touch-icon sizes=180x180 href=/images/apple-touch-icon.png><link rel=manifest href=/site.webmanifest><link rel=mask-icon href=/images/safari-pinned-tab.svg color=#5bbad5></head><body class="preload-transitions colorscheme-auto"><div class=float-container><a id=dark-mode-toggle class=colorscheme-toggle><i class="fa-solid fa-adjust fa-fw" aria-hidden=true></i></a></div><main class=wrapper><nav class=navigation><section class=container><a class=navigation-title href=/>Eric X. Liu's Personal Page
+</a><input type=checkbox id=menu-toggle>
+<label class="menu-button float-right" for=menu-toggle><i class="fa-solid fa-bars fa-fw" aria-hidden=true></i></label><ul class=navigation-list><li class=navigation-item><a class=navigation-link href=/posts/>Posts</a></li><li class=navigation-item><a class=navigation-link href=https://chat.ericxliu.me>Chat</a></li><li class=navigation-item><a class=navigation-link href=https://git.ericxliu.me/user/oauth2/Authenitk>Git</a></li><li class=navigation-item><a class=navigation-link href=https://coder.ericxliu.me/api/v2/users/oidc/callback>Coder</a></li><li class=navigation-item><a class=navigation-link href=/>|</a></li><li class=navigation-item><a class=navigation-link href=https://sso.ericxliu.me>Sign in</a></li></ul></section></nav><div class=content><section class="container post"><article><header><div class=post-title><h1 class=title><a class=title-link href=/posts/espresso-theory-application-a-guide-for-the-breville-barista-pro/>Mastering Your Breville Barista Pro: The Ultimate Guide to Dialing In Espresso</a></h1></div><div class=post-meta><div class=date><span class=posted-on><i class="fa-solid fa-calendar" aria-hidden=true></i>
+<time datetime=2025-05-01T00:00:00Z>May 1, 2025
+</time></span><span class=reading-time><i class="fa-solid fa-clock" aria-hidden=true></i>
+6-minute read</span></div></div></header><div class=post-content><p>Are you ready to transform your home espresso game from good to genuinely great? The Breville Barista Pro is a fantastic machine, but unlocking its full potential requires understanding a few key principles. This guide will walk you through the systematic process of dialing in your espresso, ensuring every shot is delicious and repeatable.</p><p>Our overarching philosophy is simple: <strong>isolate and change only one variable at a time.</strong> While numbers are crucial, your palate is the ultimate judge. Dose, ratio, and time are interconnected, but your <strong>grind size</strong> is your most powerful lever.</p><p>Let&rsquo;s dive in!</p><hr><h3 id=part-1-the-foundation--dose-the-weight-of-dry-coffee><strong>Part 1: The Foundation — Dose (The Weight of Dry Coffee)</strong>
+<a class=heading-link href=#part-1-the-foundation--dose-the-weight-of-dry-coffee><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><p>Your dose is the bedrock of your espresso. It&rsquo;s the weight of your ground coffee, and it should be the first variable you set and then keep <strong>constant</strong> during the initial dialing-in process.</p><p><strong>Why Dose Matters:</strong></p><ul><li><strong>Basket Size is Key:</strong> Your portafilter basket dictates your ideal dose. Too little coffee (under-dosing) creates excessive &ldquo;headspace,&rdquo; leading to soupy extractions. Too much (over-dosing) causes the coffee puck to touch the shower screen, preventing even water flow and causing channeling.</li><li><strong>Extraction &ldquo;Work&rdquo;:</strong> A higher dose means more coffee mass, requiring more &ldquo;work&rdquo; (a finer grind, more water) to extract properly.</li><li><strong>Coffee Type:</strong><ul><li><strong>Light Roasts:</strong> Denser and harder to extract. Consider a <strong>slightly lower dose</strong>.</li><li><strong>Dark Roasts:</strong> More brittle and soluble. You can often use a <strong>slightly higher dose</strong>.</li></ul></li></ul><p><strong>Application for Your Breville Barista Pro (54mm Portafilter):</strong></p><ul><li><strong>Your Starting Point:</strong> Always begin with <strong>18 grams</strong>. Use a scale for accuracy!</li><li><strong>Adjusting for Roast:</strong> For light roasts, if you&rsquo;re struggling, drop to 17g. For dark roasts, you can try 19g.</li><li><strong>Golden Rule:</strong> Once you choose your starting dose (e.g., 18g), <strong>do not change it</strong> until you&rsquo;ve dialed in your grind size.</li></ul><hr><h3 id=part-2-defining-the-drink--brew-ratio-dose-vs-yield><strong>Part 2: Defining the Drink — Brew Ratio (Dose vs. Yield)</strong>
+<a class=heading-link href=#part-2-defining-the-drink--brew-ratio-dose-vs-yield><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><p>The brew ratio defines the relationship between your dry coffee dose and the weight of your liquid espresso yield. Always measure by <strong>weight (grams)</strong>, not volume (mL), as crema can be inconsistent.</p><p><strong>Understanding Ratios:</strong></p><ul><li><strong>Ristretto (1:1 – 1:1.5):</strong> E.g., 18g in → 18g to 27g out. Strong, textured, less extracted.</li><li><strong>Espresso (Normale) (1:1.5 – 1:2.5):</strong> E.g., 18g in → 27g to 45g out. The standard, balanced shot.</li><li><strong>Lungo (1:2.5+):</strong> E.g., 18g in → 45g+ out. Weaker, less textured, more extracted.</li></ul><p><strong>The Fundamental Trade-Off:</strong></p><ul><li><strong>Longer Ratio (more water):</strong> Higher extraction, but lower strength (more diluted).</li><li><strong>Shorter Ratio (less water):</strong> Lower extraction, but higher strength (more concentrated).</li></ul><p><strong>Application for Your Breville Barista Pro:</strong></p><ul><li><strong>Recommended Starting Ratio:</strong> A <strong>1:2 ratio</strong> is the perfect place to begin.</li><li><strong>Practical Numbers:</strong> With your 18g dose, your target yield is <strong>36 grams</strong> of liquid espresso.</li><li><strong>Execution:</strong> Place your cup on a scale and use the manual brew function to stop the shot precisely when the scale reads 36g.</li></ul><hr><h3 id=part-3-the-diagnostic-tool--brew-time><strong>Part 3: The Diagnostic Tool — Brew Time</strong>
+<a class=heading-link href=#part-3-the-diagnostic-tool--brew-time><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><p>Brew time is not something you set directly; it&rsquo;s the <strong>result</strong> of how much resistance your coffee puck provides against the machine&rsquo;s water pressure. Think of it as a <strong>diagnostic tool</strong>.</p><p><strong>The 25-30 Second Guideline:</strong></p><p>This is a benchmark. If your 1:2 ratio shot falls within this time, your grind size is likely in the correct range for a balanced extraction.</p><ul><li><strong>Too Fast (&lt;25s):</strong> Indicates under-extraction (often tastes sour).</li><li><strong>Too Slow (>30s):</strong> Indicates over-extraction (often tastes bitter).</li></ul><p><strong>Taste is King:</strong> Remember, if a shot tastes fantastic at 32 seconds, it&rsquo;s a great shot! The time simply becomes part of your successful recipe for that specific coffee.</p><p><strong>Application for Your Breville Barista Pro:</strong></p><ul><li><strong>Pre-infusion:</strong> The Barista Pro&rsquo;s low-pressure pre-infusion is <strong>part of your total brew time</strong>. Its purpose is to saturate the puck evenly to prevent channeling. Keep it consistent for every shot while dialing in.</li></ul><hr><h3 id=part-4-the-primary-control--grind-setting><strong>Part 4: The Primary Control — Grind Setting</strong>
+<a class=heading-link href=#part-4-the-primary-control--grind-setting><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><p>This is where the magic (and sometimes frustration) happens. Grind size is your main tool for controlling the resistance of the coffee puck, which directly dictates your brew time.</p><p><strong>The Dual Impact of Grinding Finer:</strong></p><ol><li><strong>Increases surface area:</strong> Allows for more efficient flavor extraction.</li><li><strong>Increases resistance:</strong> Slows down water flow and increases contact time.</li></ol><p><strong>The Risk of Grinding Too Fine (Channeling):</strong></p><p>If the grind is too fine, the puck becomes so dense that high-pressure water can&rsquo;t flow evenly. Instead, it &ldquo;breaks&rdquo; the puck and punches an easy path (a channel) through a weak spot. This results in a disastrous shot that is simultaneously:</p><ul><li><strong>Under-extracted:</strong> Most of the coffee is bypassed.</li><li><strong>Over-extracted:</strong> The water that does flow blasts through the channel, extracting harsh, bitter compounds.</li><li><strong>The Taste:</strong> A channeled shot tastes hollow, weak, sour, <em>and</em> bitter all at once.</li></ul><p><strong>The Goal:</strong> You want to <strong>grind as fine as you possibly can <em>without</em> causing significant channeling</strong>. This is the sweet spot for maximizing surface area and resistance for high, even extraction.</p><p><strong>Grind Retention (Purging):</strong> Most grinders retain some old grounds. When you change your grind setting, always purge a few grams of coffee to ensure your dose is entirely at the new setting.</p><p><strong>Application for Your Breville Barista Pro:</strong></p><ul><li><strong>Grinder Mechanism:</strong> The &ldquo;Grind Amount&rdquo; dial controls the <strong>TIME</strong> the grinder runs, not the weight. When you adjust the fineness, you <strong>must</strong> re-adjust the grind time to ensure you are still getting your target 18g dose.</li><li><strong>Tackling Channeling:</strong> The Barista Pro is prone to channeling. To fight this, focus on excellent <strong>puck prep</strong>: use a WDT (Weiss Distribution Technique) tool to break up clumps and evenly distribute the grounds before tamping levelly.</li></ul><hr><h3 id=the-complete-dialing-in-workflow><strong>The Complete Dialing-In Workflow</strong>
+<a class=heading-link href=#the-complete-dialing-in-workflow><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><p>This systematic process will get you to a delicious shot from your Breville Barista Pro efficiently:</p><ol><li><strong>Set Your Constants:</strong><ul><li><strong>Dose:</strong> <strong>18g</strong>.</li><li><strong>Ratio:</strong> <strong>1:2</strong> (meaning a <strong>Yield</strong> of <strong>36g</strong>).</li><li><strong>Pre-infusion:</strong> Use a consistent method (e.g., manual 8-second hold).</li></ul></li><li><strong>Make an Initial Grind:</strong><ul><li>Set the grinder to a starting point of <strong>15</strong>.</li><li>Adjust the grind <strong>time</strong> until the grinder dispenses exactly 18g.</li></ul></li><li><strong>Pull the First Shot:</strong><ul><li>Brew manually, stopping at <strong>36g</strong> of liquid in the cup. Note the <strong>total brew time</strong>.</li></ul></li><li><strong>Taste and Diagnose:</strong><ul><li><strong>Fast & Sour? (&lt;25s):</strong> Grind is too coarse.</li><li><strong>Slow & Bitter? (>32s):</strong> Grind is too fine.</li></ul></li><li><strong>Make ONE Adjustment - THE GRIND SIZE:</strong><ul><li>If fast/sour, adjust the grind <strong>finer</strong> (e.g., from 15 down to 13).</li><li>If slow/bitter, adjust the grind <strong>coarser</strong> (e.g., from 15 up to 17).</li></ul></li><li><strong>Re-adjust and Repeat:</strong><ul><li>After changing the grind setting, <strong>purge</strong> a small amount of coffee.</li><li>Re-weigh your next dose and <strong>adjust the grind time</strong> to get back to exactly 18g.</li><li>Pull another 36g shot. Repeat this process until your shot tastes balanced and the time falls roughly between <strong>25-32 seconds</strong>.</li></ul></li></ol><p>Happy brewing! With patience and this systematic approach, you&rsquo;ll be pulling consistently delicious espresso shots from your Breville Barista Pro in no time.</p></div><footer><div id=disqus_thread></div><script>window.disqus_config=function(){},function(){if(["localhost","127.0.0.1"].indexOf(window.location.hostname)!=-1){document.getElementById("disqus_thread").innerHTML="Disqus comments not available by default when the website is previewed locally.";return}var t=document,e=t.createElement("script");e.async=!0,e.src="//ericxliu-me.disqus.com/embed.js",e.setAttribute("data-timestamp",+new Date),(t.head||t.body).appendChild(e)}(),document.addEventListener("themeChanged",function(){document.readyState=="complete"&&DISQUS.reset({reload:!0,config:disqus_config})})</script></footer></article><link rel=stylesheet href=https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/katex.min.css integrity=sha384-vKruj+a13U8yHIkAyGgK1J3ArTLzrFGBbBc0tDp4ad/EyewESeXE/Iv67Aj8gKZ0 crossorigin=anonymous><script defer src=https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/katex.min.js integrity=sha384-PwRUT/YqbnEjkZO0zZxNqcxACrXe+j766U2amXcgMg5457rve2Y7I6ZJSm2A0mS4 crossorigin=anonymous></script><script defer src=https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/contrib/auto-render.min.js integrity=sha384-+VBxd3r6XgURycqtZ117nYw44OOcIax56Z4dCRWbxyPt0Koah1uHoK0o4+/RRE05 crossorigin=anonymous onload='renderMathInElement(document.body,{delimiters:[{left:"$$",right:"$$",display:!0},{left:"$",right:"$",display:!1},{left:"\\(",right:"\\)",display:!1},{left:"\\[",right:"\\]",display:!0}]})'></script></section></div><footer class=footer><section class=container>©
+2016 -
+2025
+Eric X. Liu
+<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/69c2890">[69c2890]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script><script defer src=https://static.cloudflareinsights.com/beacon.min.js data-cf-beacon='{"token": "987638e636ce4dbb932d038af74c17d1"}'></script></body></html>
--- a/posts/how-rvq-teaches-llms-to-see-and-hear/index.html
+++ b/posts/how-rvq-teaches-llms-to-see-and-hear/index.html
@@ -0,0 +1,21 @@
+<!doctype html><html lang=en><head><title>Beyond Words: How RVQ Teaches LLMs to See and Hear · Eric X. Liu's Personal Page</title><meta charset=utf-8><meta name=viewport content="width=device-width,initial-scale=1"><meta name=color-scheme content="light dark"><meta name=author content="Eric X. Liu"><meta name=description content="Large Language Models (LLMs) are masters of text, but the world is not made of text alone. It’s a symphony of sights, sounds, and experiences. The ultimate goal for AI is to understand this rich, multi-modal world as we do. But how do you teach a model that thinks in words to understand a picture of a sunset or the melody of a song?
+The answer lies in creating a universal language—a bridge between the continuous, messy world of pixels and audio waves and the discrete, structured world of language tokens. One of the most elegant and powerful tools for building this bridge is Residual Vector Quantization (RVQ)."><meta name=keywords content="software engineer,performance engineering,Google engineer,tech blog,software development,performance optimization,Eric Liu,engineering blog,mountain biking,Jeep enthusiast,overlanding,camping,outdoor adventures"><meta name=twitter:card content="summary"><meta name=twitter:title content="Beyond Words: How RVQ Teaches LLMs to See and Hear"><meta name=twitter:description content="Large Language Models (LLMs) are masters of text, but the world is not made of text alone. It’s a symphony of sights, sounds, and experiences. The ultimate goal for AI is to understand this rich, multi-modal world as we do. But how do you teach a model that thinks in words to understand a picture of a sunset or the melody of a song?
+The answer lies in creating a universal language—a bridge between the continuous, messy world of pixels and audio waves and the discrete, structured world of language tokens. One of the most elegant and powerful tools for building this bridge is Residual Vector Quantization (RVQ)."><meta property="og:url" content="/posts/how-rvq-teaches-llms-to-see-and-hear/"><meta property="og:site_name" content="Eric X. Liu's Personal Page"><meta property="og:title" content="Beyond Words: How RVQ Teaches LLMs to See and Hear"><meta property="og:description" content="Large Language Models (LLMs) are masters of text, but the world is not made of text alone. It’s a symphony of sights, sounds, and experiences. The ultimate goal for AI is to understand this rich, multi-modal world as we do. But how do you teach a model that thinks in words to understand a picture of a sunset or the melody of a song?
+The answer lies in creating a universal language—a bridge between the continuous, messy world of pixels and audio waves and the discrete, structured world of language tokens. One of the most elegant and powerful tools for building this bridge is Residual Vector Quantization (RVQ)."><meta property="og:locale" content="en"><meta property="og:type" content="article"><meta property="article:section" content="posts"><meta property="article:published_time" content="2025-08-07T00:00:00+00:00"><meta property="article:modified_time" content="2025-08-08T17:36:52+00:00"><link rel=canonical href=/posts/how-rvq-teaches-llms-to-see-and-hear/><link rel=preload href=/fonts/fa-brands-400.woff2 as=font type=font/woff2 crossorigin><link rel=preload href=/fonts/fa-regular-400.woff2 as=font type=font/woff2 crossorigin><link rel=preload href=/fonts/fa-solid-900.woff2 as=font type=font/woff2 crossorigin><link rel=stylesheet href=/css/coder.min.6445a802b9389c9660e1b07b724dcf5718b1065ed2d71b4eeaf981cc7cc5fc46.css integrity="sha256-ZEWoArk4nJZg4bB7ck3PVxixBl7S1xtO6vmBzHzF/EY=" crossorigin=anonymous media=screen><link rel=stylesheet href=/css/coder-dark.min.a00e6364bacbc8266ad1cc81230774a1397198f8cfb7bcba29b7d6fcb54ce57f.css integrity="sha256-oA5jZLrLyCZq0cyBIwd0oTlxmPjPt7y6KbfW/LVM5X8=" crossorigin=anonymous media=screen><link rel=icon type=image/svg+xml href=/images/favicon.svg sizes=any><link rel=icon type=image/png href=/images/favicon-32x32.png sizes=32x32><link rel=icon type=image/png href=/images/favicon-16x16.png sizes=16x16><link rel=apple-touch-icon href=/images/apple-touch-icon.png><link rel=apple-touch-icon sizes=180x180 href=/images/apple-touch-icon.png><link rel=manifest href=/site.webmanifest><link rel=mask-icon href=/images/safari-pinned-tab.svg color=#5bbad5></head><body class="preload-transitions colorscheme-auto"><div class=float-container><a id=dark-mode-toggle class=colorscheme-toggle><i class="fa-solid fa-adjust fa-fw" aria-hidden=true></i></a></div><main class=wrapper><nav class=navigation><section class=container><a class=navigation-title href=/>Eric X. Liu's Personal Page
+</a><input type=checkbox id=menu-toggle>
+<label class="menu-button float-right" for=menu-toggle><i class="fa-solid fa-bars fa-fw" aria-hidden=true></i></label><ul class=navigation-list><li class=navigation-item><a class=navigation-link href=/posts/>Posts</a></li><li class=navigation-item><a class=navigation-link href=https://chat.ericxliu.me>Chat</a></li><li class=navigation-item><a class=navigation-link href=https://git.ericxliu.me/user/oauth2/Authenitk>Git</a></li><li class=navigation-item><a class=navigation-link href=https://coder.ericxliu.me/api/v2/users/oidc/callback>Coder</a></li><li class=navigation-item><a class=navigation-link href=/>|</a></li><li class=navigation-item><a class=navigation-link href=https://sso.ericxliu.me>Sign in</a></li></ul></section></nav><div class=content><section class="container post"><article><header><div class=post-title><h1 class=title><a class=title-link href=/posts/how-rvq-teaches-llms-to-see-and-hear/>Beyond Words: How RVQ Teaches LLMs to See and Hear</a></h1></div><div class=post-meta><div class=date><span class=posted-on><i class="fa-solid fa-calendar" aria-hidden=true></i>
+<time datetime=2025-08-07T00:00:00Z>August 7, 2025
+</time></span><span class=reading-time><i class="fa-solid fa-clock" aria-hidden=true></i>
+6-minute read</span></div></div></header><div class=post-content><p>Large Language Models (LLMs) are masters of text, but the world is not made of text alone. It’s a symphony of sights, sounds, and experiences. The ultimate goal for AI is to understand this rich, multi-modal world as we do. But how do you teach a model that thinks in words to understand a picture of a sunset or the melody of a song?</p><p>The answer lies in creating a universal language—a bridge between the continuous, messy world of pixels and audio waves and the discrete, structured world of language tokens. One of the most elegant and powerful tools for building this bridge is <strong>Residual Vector Quantization (RVQ)</strong>.</p><p>This article dives deep into RVQ, exploring how it turns raw data into meaningful semantic IDs and how these IDs, in turn, unlock multi-modal understanding in LLMs.</p><h4 id=what-is-residual-vector-quantization-the-art-of-smart-compression><strong>What is Residual Vector Quantization? The Art of Smart Compression</strong>
+<a class=heading-link href=#what-is-residual-vector-quantization-the-art-of-smart-compression><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h4><p>At its core, Vector Quantization (VQ) is a compression technique. It maps a high-dimensional vector (like an data embedding) to the single closest vector in a predefined dictionary, called a <strong>codebook</strong>. You then only need to store the index of that chosen vector. The problem? To represent complex data accurately, you&rsquo;d need a codebook with an astronomical number of entries, which is computationally impossible.</p><p>This is where <strong>Residual</strong> Vector Quantization shines. Instead of one giant codebook, RVQ uses a series of smaller codebooks in stages.</p><ol><li><strong>Stage 1 (Coarse Quantization):</strong> The input vector is quantized by the first codebook. This finds the broadest, most general category for the data.</li><li><strong>Calculate the Residual:</strong> The system calculates the error, or &ldquo;residual,&rdquo; between the original vector and its quantized version from Stage 1. This residual vector represents the information that was lost in the first coarse approximation.</li><li><strong>Stage 2 (Refinement):</strong> This residual vector is then quantized by the <em>second</em> codebook. This stage doesn&rsquo;t re-evaluate the whole vector, but only focuses on correcting the error from the previous stage.</li><li><strong>Iterate:</strong> This process repeats for several stages, with each subsequent codebook quantizing the residual error from the previous one, adding a finer and finer layer of detail.</li></ol><p>The final compressed representation is simply the sequence of indices from each codebook. For example, an ID like <code>[8, 5, 4, 1]</code> is produced. The magic of this approach is that it creates a <strong>hierarchical ID</strong>. The first digit <code>[8]</code> might represent &ldquo;Sports,&rdquo; the next <code>[5]</code> refines it to &ldquo;Court Sports,&rdquo; <code>[4]</code> to &ldquo;Beach Volleyball,&rdquo; and the final <code>[1]</code> distinguishes a specific match. Videos with similar content will naturally share a longer prefix in their Semantic ID.</p><h4 id=learning-what-matters-the-trainable-vq-autoencoder><strong>Learning What Matters: The Trainable VQ-Autoencoder</strong>
+<a class=heading-link href=#learning-what-matters-the-trainable-vq-autoencoder><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h4><p>A key insight is that RVQ is not a fixed algorithm but a <strong>trainable neural network component</strong>. Its codebooks are not predefined; they are learned. This learning happens within a <strong>Vector-Quantized Autoencoder (VQ-AE)</strong> architecture.</p><ol><li><strong>Encoder:</strong> A powerful neural network (e.g., a Transformer or CNN) takes the raw data (like video frames and audio) and converts it into a continuous semantic embedding.</li><li><strong>RVQ Bottleneck:</strong> This embedding is fed into the RVQ module, which quantizes it into the sequence of discrete IDs.</li><li><strong>Decoder:</strong> The decoder takes these discrete IDs, looks up the corresponding codebook vectors, sums them up to get a reconstructed embedding, and attempts to rebuild the original video/audio.</li></ol><p>The entire system is trained end-to-end. The <strong>reconstruction loss</strong> (the difference between the original and reconstructed data) is used to update the parameters of the Encoder, the Decoder, and, most importantly, <strong>the codebook vectors within the RVQ module</strong>. Initially random, the codebook vectors are gradually pushed to become meaningful &ldquo;anchors&rdquo; for the core concepts present in the training data.</p><h4 id=from-implicit-to-explicit-controlling-semantics-with-contrastive-learning><strong>From Implicit to Explicit: Controlling Semantics with Contrastive Learning</strong>
+<a class=heading-link href=#from-implicit-to-explicit-controlling-semantics-with-contrastive-learning><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h4><p>A standard VQ-AE learns implicit semantics. It gets good at reconstruction, but we can&rsquo;t control <em>what</em> concepts it learns. To make the Semantic IDs truly meaningful and aligned with human language, we introduce <strong>contrastive learning</strong>.</p><p>The architecture is enhanced with a parallel text encoder (like BERT or CLIP&rsquo;s). The model is then trained with a joint loss function:</p><p><code>L_total = L_reconstruction + λ * L_contrastive</code></p><ul><li><strong>Reconstruction Loss</strong> ensures the RVQ codes contain enough information to rebuild the input.</li><li><strong>Contrastive Loss</strong> forces the media embedding (from the video/audio encoder) to be mathematically &ldquo;close&rdquo; to the text embedding of its description, and &ldquo;far&rdquo; from the embeddings of unrelated text descriptions.</li></ul><p>This dual goal forces the model to organize its embedding space according to the semantics of human language. The codebook vectors now learn to represent concepts that are not just useful for reconstruction, but are also tied to explicit textual descriptions.</p><h4 id=integrating-with-llms-two-powerful-paths-to-multi-modality><strong>Integrating with LLMs: Two Powerful Paths to Multi-Modality</strong>
+<a class=heading-link href=#integrating-with-llms-two-powerful-paths-to-multi-modality><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h4><p>Once we have a contrastively-trained VQ-AE, we can use its output to give LLMs the ability to see and hear. There are two primary strategies for this.</p><p><strong>Path 1: The Tokenizer Approach - Teaching the LLM a New Language</strong></p><p>This path treats the RVQ IDs as a new vocabulary. It’s a two-stage process ideal for high-fidelity content generation.</p><ol><li><strong>Create a Neural Codec:</strong> The trained VQ-AE serves as a powerful &ldquo;codec.&rdquo; You can take any piece of media (e.g., a song) and use the codec to compress it into a sequence of discrete RVQ tokens (e.g., <code>[8, 5, 4, 1, 8, 5, 9, 2, ...]</code>).</li><li><strong>Train a Generative LLM:</strong> A new Transformer model is trained auto-regressively on a massive dataset of these media-derived tokens. Its sole purpose is to learn the patterns and predict the next token in a sequence.</li></ol><p><strong>Use Case:</strong> This is the architecture behind models like Meta&rsquo;s MusicGen. A user provides a text prompt, which conditions the Transformer to generate a new sequence of RVQ tokens. These tokens are then fed to the VQ-AE&rsquo;s decoder to synthesize the final audio waveform.</p><p><strong>Path 2: The Adapter Approach - Translating for a Language Expert</strong></p><p>This path is used to augment a powerful, pre-trained, text-only LLM without the astronomical cost of retraining it.</p><ol><li><strong>Freeze the LLM:</strong> A massive, pre-trained LLM (like LLaMA) is frozen. Its deep language understanding is preserved.</li><li><strong>Use the Pre-Quantized Embedding:</strong> Instead of using the discrete RVQ tokens, we take the rich, continuous embedding vector produced by our media encoder <em>just before</em> it enters the RVQ module.</li><li><strong>Train a Small Adapter:</strong> A small, lightweight projection layer (or &ldquo;adapter&rdquo;) is trained. Its only job is to translate the media embedding into a vector that has the same format and structure as the LLM&rsquo;s own word embeddings. It learns to map visual concepts to their corresponding &ldquo;word&rdquo; concepts in the LLM&rsquo;s latent space.</li></ol><p><strong>Use Case:</strong> This is the principle behind models like Google&rsquo;s Flamingo. To answer a question about an image, the image is passed through the media encoder and adapter. The resulting &ldquo;vision-as-a-word&rdquo; vector is inserted into the prompt sequence alongside the text tokens. The frozen LLM can now &ldquo;reason&rdquo; about the visual input because it has been translated into a format it already understands.</p></div><footer><div id=disqus_thread></div><script>window.disqus_config=function(){},function(){if(["localhost","127.0.0.1"].indexOf(window.location.hostname)!=-1){document.getElementById("disqus_thread").innerHTML="Disqus comments not available by default when the website is previewed locally.";return}var t=document,e=t.createElement("script");e.async=!0,e.src="//ericxliu-me.disqus.com/embed.js",e.setAttribute("data-timestamp",+new Date),(t.head||t.body).appendChild(e)}(),document.addEventListener("themeChanged",function(){document.readyState=="complete"&&DISQUS.reset({reload:!0,config:disqus_config})})</script></footer></article><link rel=stylesheet href=https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/katex.min.css integrity=sha384-vKruj+a13U8yHIkAyGgK1J3ArTLzrFGBbBc0tDp4ad/EyewESeXE/Iv67Aj8gKZ0 crossorigin=anonymous><script defer src=https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/katex.min.js integrity=sha384-PwRUT/YqbnEjkZO0zZxNqcxACrXe+j766U2amXcgMg5457rve2Y7I6ZJSm2A0mS4 crossorigin=anonymous></script><script defer src=https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/contrib/auto-render.min.js integrity=sha384-+VBxd3r6XgURycqtZ117nYw44OOcIax56Z4dCRWbxyPt0Koah1uHoK0o4+/RRE05 crossorigin=anonymous onload='renderMathInElement(document.body,{delimiters:[{left:"$$",right:"$$",display:!0},{left:"$",right:"$",display:!1},{left:"\\(",right:"\\)",display:!1},{left:"\\[",right:"\\]",display:!0}]})'></script></section></div><footer class=footer><section class=container>©
+2016 -
+2025
+Eric X. Liu
+<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/69c2890">[69c2890]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script><script defer src=https://static.cloudflareinsights.com/beacon.min.js data-cf-beacon='{"token": "987638e636ce4dbb932d038af74c17d1"}'></script></body></html>
--- a/posts/index.html
+++ b/posts/index.html
@@ -0,0 +1,17 @@
+<!doctype html><html lang=en><head><title>Posts · Eric X. Liu's Personal Page</title><meta charset=utf-8><meta name=viewport content="width=device-width,initial-scale=1"><meta name=color-scheme content="light dark"><meta name=author content="Eric X. Liu"><meta name=description content="Eric X. Liu - Software & Performance Engineer at Google. Sharing insights about software engineering, performance optimization, tech industry experiences, mountain biking adventures, Jeep overlanding, and outdoor activities."><meta name=keywords content="software engineer,performance engineering,Google engineer,tech blog,software development,performance optimization,Eric Liu,engineering blog,mountain biking,Jeep enthusiast,overlanding,camping,outdoor adventures"><meta name=twitter:card content="summary"><meta name=twitter:title content="Posts"><meta name=twitter:description content="Eric X. Liu - Software & Performance Engineer at Google. Sharing insights about software engineering, performance optimization, tech industry experiences, mountain biking adventures, Jeep overlanding, and outdoor activities."><meta property="og:url" content="/posts/"><meta property="og:site_name" content="Eric X. Liu's Personal Page"><meta property="og:title" content="Posts"><meta property="og:description" content="Eric X. Liu - Software & Performance Engineer at Google. Sharing insights about software engineering, performance optimization, tech industry experiences, mountain biking adventures, Jeep overlanding, and outdoor activities."><meta property="og:locale" content="en"><meta property="og:type" content="website"><link rel=canonical href=/posts/><link rel=preload href=/fonts/fa-brands-400.woff2 as=font type=font/woff2 crossorigin><link rel=preload href=/fonts/fa-regular-400.woff2 as=font type=font/woff2 crossorigin><link rel=preload href=/fonts/fa-solid-900.woff2 as=font type=font/woff2 crossorigin><link rel=stylesheet href=/css/coder.min.6445a802b9389c9660e1b07b724dcf5718b1065ed2d71b4eeaf981cc7cc5fc46.css integrity="sha256-ZEWoArk4nJZg4bB7ck3PVxixBl7S1xtO6vmBzHzF/EY=" crossorigin=anonymous media=screen><link rel=stylesheet href=/css/coder-dark.min.a00e6364bacbc8266ad1cc81230774a1397198f8cfb7bcba29b7d6fcb54ce57f.css integrity="sha256-oA5jZLrLyCZq0cyBIwd0oTlxmPjPt7y6KbfW/LVM5X8=" crossorigin=anonymous media=screen><link rel=icon type=image/svg+xml href=/images/favicon.svg sizes=any><link rel=icon type=image/png href=/images/favicon-32x32.png sizes=32x32><link rel=icon type=image/png href=/images/favicon-16x16.png sizes=16x16><link rel=apple-touch-icon href=/images/apple-touch-icon.png><link rel=apple-touch-icon sizes=180x180 href=/images/apple-touch-icon.png><link rel=manifest href=/site.webmanifest><link rel=mask-icon href=/images/safari-pinned-tab.svg color=#5bbad5><link rel=alternate type=application/rss+xml href=/posts/index.xml title="Eric X. Liu's Personal Page"></head><body class="preload-transitions colorscheme-auto"><div class=float-container><a id=dark-mode-toggle class=colorscheme-toggle><i class="fa-solid fa-adjust fa-fw" aria-hidden=true></i></a></div><main class=wrapper><nav class=navigation><section class=container><a class=navigation-title href=/>Eric X. Liu's Personal Page
+</a><input type=checkbox id=menu-toggle>
+<label class="menu-button float-right" for=menu-toggle><i class="fa-solid fa-bars fa-fw" aria-hidden=true></i></label><ul class=navigation-list><li class=navigation-item><a class=navigation-link href=/posts/>Posts</a></li><li class=navigation-item><a class=navigation-link href=https://chat.ericxliu.me>Chat</a></li><li class=navigation-item><a class=navigation-link href=https://git.ericxliu.me/user/oauth2/Authenitk>Git</a></li><li class=navigation-item><a class=navigation-link href=https://coder.ericxliu.me/api/v2/users/oidc/callback>Coder</a></li><li class=navigation-item><a class=navigation-link href=/>|</a></li><li class=navigation-item><a class=navigation-link href=https://sso.ericxliu.me>Sign in</a></li></ul></section></nav><div class=content><section class="container list"><header><h1 class=title><a class=title-link href=/posts/>Posts</a></h1></header><ul><li><span class=date>August 19, 2025</span>
+<a class=title href=/posts/quantization-in-llms/>Quantization in LLMs</a></li><li><span class=date>August 16, 2025</span>
+<a class=title href=/posts/breville-barista-pro-maintenance/>Breville Barista Pro Maintenance</a></li><li><span class=date>August 9, 2025</span>
+<a class=title href=/posts/secure-boot-dkms-and-mok-on-proxmox-debian/>Fixing GPU Operator Pods Stuck in Init: Secure Boot, DKMS, and MOK on Proxmox + Debian</a></li><li><span class=date>August 7, 2025</span>
+<a class=title href=/posts/how-rvq-teaches-llms-to-see-and-hear/>Beyond Words: How RVQ Teaches LLMs to See and Hear</a></li><li><span class=date>August 3, 2025</span>
+<a class=title href=/posts/supabase-deep-dive/>Supabase Deep Dive: It's Not Magic, It's Just Postgres</a></li><li><span class=date>August 2, 2025</span>
+<a class=title href=/posts/ppo-for-language-models/>A Deep Dive into PPO for Language Models</a></li><li><span class=date>July 2, 2025</span>
+<a class=title href=/posts/mixture-of-experts-moe-models-challenges-solutions-in-practice/>Mixture-of-Experts (MoE) Models Challenges & Solutions in Practice</a></li><li><span class=date>June 1, 2025</span>
+<a class=title href=/posts/t5-the-transformer-that-zigged-when-others-zagged-an-architectural-deep-dive/>An Architectural Deep Dive of T5</a></li><li><span class=date>May 1, 2025</span>
+<a class=title href=/posts/espresso-theory-application-a-guide-for-the-breville-barista-pro/>Mastering Your Breville Barista Pro: The Ultimate Guide to Dialing In Espresso</a></li><li><span class=date>April 1, 2025</span>
+<a class=title href=/posts/transformer-s-core-mechanics/>Transformer's Core Mechanics</a></li></ul><ul class=pagination><li>1</li><li><a href=/posts/page/2/>2</a></li><li class=hidden><a href=/posts/page/2/>&#8250;</a></li><li><a href=/posts/page/2/>&#187;</a></li></ul></section></div><footer class=footer><section class=container>©
+2016 -
+2025
+Eric X. Liu
+<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/69c2890">[69c2890]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script><script defer src=https://static.cloudflareinsights.com/beacon.min.js data-cf-beacon='{"token": "987638e636ce4dbb932d038af74c17d1"}'></script></body></html>
--- a/posts/index.xml
+++ b/posts/index.xml
@@ -0,0 +1,38 @@
+<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Posts on Eric X. Liu's Personal Page</title><link>/posts/</link><description>Recent content in Posts on Eric X. Liu's Personal Page</description><generator>Hugo</generator><language>en</language><lastBuildDate>Wed, 20 Aug 2025 06:28:39 +0000</lastBuildDate><atom:link href="/posts/index.xml" rel="self" type="application/rss+xml"/><item><title>Quantization in LLMs</title><link>/posts/quantization-in-llms/</link><pubDate>Tue, 19 Aug 2025 00:00:00 +0000</pubDate><guid>/posts/quantization-in-llms/</guid><description>&lt;p>The burgeoning scale of Large Language Models (LLMs) has necessitated a paradigm shift in their deployment, moving beyond full-precision floating-point arithmetic towards lower-precision representations. Quantization, the process of mapping a wide range of continuous values to a smaller, discrete set, has emerged as a critical technique to reduce model size, accelerate inference, and lower energy consumption. This article provides a technical overview of quantization theories, their application in modern LLMs, and highlights the ongoing innovations in this domain.&lt;/p></description></item><item><title>Breville Barista Pro Maintenance</title><link>/posts/breville-barista-pro-maintenance/</link><pubDate>Sat, 16 Aug 2025 00:00:00 +0000</pubDate><guid>/posts/breville-barista-pro-maintenance/</guid><description>&lt;p>Proper maintenance is critical for the longevity and performance of a Breville Barista Pro espresso machine. Consistent cleaning not only ensures the machine functions correctly but also directly impacts the quality of the espresso produced. This guide provides a detailed, technical breakdown of the essential maintenance routines, from automated cycles to daily upkeep.&lt;/p>
+&lt;h4 id="understanding-the-two-primary-maintenance-cycles">
+ &lt;strong>Understanding the Two Primary Maintenance Cycles&lt;/strong>
+ &lt;a class="heading-link" href="#understanding-the-two-primary-maintenance-cycles">
+ &lt;i class="fa-solid fa-link" aria-hidden="true" title="Link to heading">&lt;/i>
+ &lt;span class="sr-only">Link to heading&lt;/span>
+ &lt;/a>
+&lt;/h4>
+&lt;p>The Breville Barista Pro has two distinct, automated maintenance procedures: the &lt;strong>Cleaning (Flush) Cycle&lt;/strong> and the &lt;strong>Descale Cycle&lt;/strong>. It is important to understand that these are not interchangeable, as they address different types of buildup within the machine.&lt;/p></description></item><item><title>Fixing GPU Operator Pods Stuck in Init: Secure Boot, DKMS, and MOK on Proxmox + Debian</title><link>/posts/secure-boot-dkms-and-mok-on-proxmox-debian/</link><pubDate>Sat, 09 Aug 2025 00:00:00 +0000</pubDate><guid>/posts/secure-boot-dkms-and-mok-on-proxmox-debian/</guid><description>&lt;p>I hit an issue where all GPU Operator pods on one node were stuck in Init after migrating from Legacy BIOS to UEFI. The common error was NVIDIA components waiting for “toolkit-ready,” while the toolkit init container looped with:&lt;/p>
+&lt;ul>
+&lt;li>nvidia-smi failed to communicate with the NVIDIA driver&lt;/li>
+&lt;li>modprobe nvidia → “Key was rejected by service”&lt;/li>
+&lt;/ul>
+&lt;p>That message is the tell: Secure Boot is enabled and the kernel refuses to load modules not signed by a trusted key.&lt;/p></description></item><item><title>Beyond Words: How RVQ Teaches LLMs to See and Hear</title><link>/posts/how-rvq-teaches-llms-to-see-and-hear/</link><pubDate>Thu, 07 Aug 2025 00:00:00 +0000</pubDate><guid>/posts/how-rvq-teaches-llms-to-see-and-hear/</guid><description>&lt;p>Large Language Models (LLMs) are masters of text, but the world is not made of text alone. It’s a symphony of sights, sounds, and experiences. The ultimate goal for AI is to understand this rich, multi-modal world as we do. But how do you teach a model that thinks in words to understand a picture of a sunset or the melody of a song?&lt;/p>
+&lt;p>The answer lies in creating a universal language—a bridge between the continuous, messy world of pixels and audio waves and the discrete, structured world of language tokens. One of the most elegant and powerful tools for building this bridge is &lt;strong>Residual Vector Quantization (RVQ)&lt;/strong>.&lt;/p></description></item><item><title>Supabase Deep Dive: It's Not Magic, It's Just Postgres</title><link>/posts/supabase-deep-dive/</link><pubDate>Sun, 03 Aug 2025 00:00:00 +0000</pubDate><guid>/posts/supabase-deep-dive/</guid><description>&lt;p>In the world of Backend-as-a-Service (BaaS), platforms are often treated as magic boxes. You push data in, you get data out, and you hope the magic inside scales. While this simplicity is powerful, it can obscure the underlying mechanics, leaving developers wondering what&amp;rsquo;s really going on.&lt;/p>
+&lt;p>Supabase enters this space with a radically different philosophy: &lt;strong>transparency&lt;/strong>. It provides the convenience of a BaaS, but it’s built on the world&amp;rsquo;s most trusted relational database: PostgreSQL. The &amp;ldquo;magic&amp;rdquo; isn&amp;rsquo;t a proprietary black box; it&amp;rsquo;s a carefully assembled suite of open-source tools that enhance Postgres, not hide it.&lt;/p></description></item><item><title>A Deep Dive into PPO for Language Models</title><link>/posts/ppo-for-language-models/</link><pubDate>Sat, 02 Aug 2025 00:00:00 +0000</pubDate><guid>/posts/ppo-for-language-models/</guid><description>&lt;p>Large Language Models (LLMs) have demonstrated astonishing capabilities, but out-of-the-box, they are simply powerful text predictors. They don&amp;rsquo;t inherently understand what makes a response helpful, harmless, or aligned with human values. The technique that has proven most effective at bridging this gap is Reinforcement Learning from Human Feedback (RLHF), and at its heart lies a powerful algorithm: Proximal Policy Optimization (PPO).&lt;/p>
+&lt;p>You may have seen diagrams like the one below, which outlines the RLHF training process. It can look intimidating, with a web of interconnected models, losses, and data flows.&lt;/p></description></item><item><title>Mixture-of-Experts (MoE) Models Challenges &amp; Solutions in Practice</title><link>/posts/mixture-of-experts-moe-models-challenges-solutions-in-practice/</link><pubDate>Wed, 02 Jul 2025 00:00:00 +0000</pubDate><guid>/posts/mixture-of-experts-moe-models-challenges-solutions-in-practice/</guid><description>&lt;p>Mixture-of-Experts (MoEs) are neural network architectures that allow different parts of the model (called &amp;ldquo;experts&amp;rdquo;) to specialize in different types of inputs. A &amp;ldquo;gating network&amp;rdquo; or &amp;ldquo;router&amp;rdquo; learns to dispatch each input (or &amp;ldquo;token&amp;rdquo;) to a subset of these experts. While powerful for scaling models, MoEs introduce several practical challenges.&lt;/p>
+&lt;h3 id="1-challenge-non-differentiability-of-routing-functions">
+ 1. Challenge: Non-Differentiability of Routing Functions
+ &lt;a class="heading-link" href="#1-challenge-non-differentiability-of-routing-functions">
+ &lt;i class="fa-solid fa-link" aria-hidden="true" title="Link to heading">&lt;/i>
+ &lt;span class="sr-only">Link to heading&lt;/span>
+ &lt;/a>
+&lt;/h3>
+&lt;p>&lt;strong>The Problem:&lt;/strong>
+Many routing mechanisms, especially &amp;ldquo;Top-K routing,&amp;rdquo; involve a discrete, hard selection process. A common function is &lt;code>KeepTopK(v, k)&lt;/code>, which selects the top &lt;code>k&lt;/code> scoring elements from a vector &lt;code>v&lt;/code> and sets others to $-\infty$ or $0$.&lt;/p></description></item><item><title>An Architectural Deep Dive of T5</title><link>/posts/t5-the-transformer-that-zigged-when-others-zagged-an-architectural-deep-dive/</link><pubDate>Sun, 01 Jun 2025 00:00:00 +0000</pubDate><guid>/posts/t5-the-transformer-that-zigged-when-others-zagged-an-architectural-deep-dive/</guid><description>&lt;p>In the rapidly evolving landscape of Large Language Models, a few key architectures define the dominant paradigms. Today, the &amp;ldquo;decoder-only&amp;rdquo; model, popularized by the GPT series and its successors like LLaMA and Mistral, reigns supreme. These models are scaled to incredible sizes and excel at in-context learning.&lt;/p>
+&lt;p>But to truly understand the field, we must look at the pivotal models that explored different paths. Google&amp;rsquo;s T5, or &lt;strong>Text-to-Text Transfer Transformer&lt;/strong>, stands out as one of the most influential. It didn&amp;rsquo;t just introduce a new model; it proposed a new philosophy. This article dives deep into the architecture of T5, how it fundamentally differs from modern LLMs, and the lasting legacy of its unique design choices.&lt;/p></description></item><item><title>Mastering Your Breville Barista Pro: The Ultimate Guide to Dialing In Espresso</title><link>/posts/espresso-theory-application-a-guide-for-the-breville-barista-pro/</link><pubDate>Thu, 01 May 2025 00:00:00 +0000</pubDate><guid>/posts/espresso-theory-application-a-guide-for-the-breville-barista-pro/</guid><description>&lt;p>Are you ready to transform your home espresso game from good to genuinely great? The Breville Barista Pro is a fantastic machine, but unlocking its full potential requires understanding a few key principles. This guide will walk you through the systematic process of dialing in your espresso, ensuring every shot is delicious and repeatable.&lt;/p>
+&lt;p>Our overarching philosophy is simple: &lt;strong>isolate and change only one variable at a time.&lt;/strong> While numbers are crucial, your palate is the ultimate judge. Dose, ratio, and time are interconnected, but your &lt;strong>grind size&lt;/strong> is your most powerful lever.&lt;/p></description></item><item><title>Transformer's Core Mechanics</title><link>/posts/transformer-s-core-mechanics/</link><pubDate>Tue, 01 Apr 2025 00:00:00 +0000</pubDate><guid>/posts/transformer-s-core-mechanics/</guid><description>&lt;p>The Transformer architecture is the bedrock of modern Large Language Models (LLMs). While its high-level success is widely known, a deeper understanding requires dissecting its core components. This article provides a detailed, technical breakdown of the fundamental concepts within a Transformer block, from the notion of &amp;ldquo;channels&amp;rdquo; to the intricate workings of the attention mechanism and its relationship with other advanced architectures like Mixture of Experts.&lt;/p>
+&lt;h3 id="1-the-channel-a-foundational-view-of-d_model">
+ 1. The &amp;ldquo;Channel&amp;rdquo;: A Foundational View of &lt;code>d_model&lt;/code>
+ &lt;a class="heading-link" href="#1-the-channel-a-foundational-view-of-d_model">
+ &lt;i class="fa-solid fa-link" aria-hidden="true" title="Link to heading">&lt;/i>
+ &lt;span class="sr-only">Link to heading&lt;/span>
+ &lt;/a>
+&lt;/h3>
+&lt;p>In deep learning, a &amp;ldquo;channel&amp;rdquo; can be thought of as a feature dimension. While this term is common in Convolutional Neural Networks for images (e.g., Red, Green, Blue channels), in LLMs, the analogous concept is the model&amp;rsquo;s primary embedding dimension, commonly referred to as &lt;code>d_model&lt;/code>.&lt;/p></description></item><item><title>Some useful files</title><link>/posts/useful/</link><pubDate>Mon, 26 Oct 2020 04:14:43 +0000</pubDate><guid>/posts/useful/</guid><description>&lt;ul>
+&lt;li>&lt;a href="/rootCA.crt" >rootCA.pem&lt;/a>&lt;/li>
+&lt;/ul></description></item></channel></rss>
--- a/posts/mixture-of-experts-moe-models-challenges-solutions-in-practice/index.html
+++ b/posts/mixture-of-experts-moe-models-challenges-solutions-in-practice/index.html
@@ -0,0 +1,47 @@
+<!doctype html><html lang=en><head><title>Mixture-of-Experts (MoE) Models Challenges & Solutions in Practice · Eric X. Liu's Personal Page</title><meta charset=utf-8><meta name=viewport content="width=device-width,initial-scale=1"><meta name=color-scheme content="light dark"><meta name=author content="Eric X. Liu"><meta name=description content="Mixture-of-Experts (MoEs) are neural network architectures that allow different parts of the model (called &ldquo;experts&rdquo;) to specialize in different types of inputs. A &ldquo;gating network&rdquo; or &ldquo;router&rdquo; learns to dispatch each input (or &ldquo;token&rdquo;) to a subset of these experts. While powerful for scaling models, MoEs introduce several practical challenges.
+
+  1. Challenge: Non-Differentiability of Routing Functions
+  
+    
+    Link to heading
+  
+
+The Problem:
+Many routing mechanisms, especially &ldquo;Top-K routing,&rdquo; involve a discrete, hard selection process. A common function is KeepTopK(v, k), which selects the top k scoring elements from a vector v and sets others to $-\infty$ or $0$."><meta name=keywords content="software engineer,performance engineering,Google engineer,tech blog,software development,performance optimization,Eric Liu,engineering blog,mountain biking,Jeep enthusiast,overlanding,camping,outdoor adventures"><meta name=twitter:card content="summary"><meta name=twitter:title content="Mixture-of-Experts (MoE) Models Challenges & Solutions in Practice"><meta name=twitter:description content="Mixture-of-Experts (MoEs) are neural network architectures that allow different parts of the model (called “experts”) to specialize in different types of inputs. A “gating network” or “router” learns to dispatch each input (or “token”) to a subset of these experts. While powerful for scaling models, MoEs introduce several practical challenges.
+1. Challenge: Non-Differentiability of Routing Functions Link to heading The Problem: Many routing mechanisms, especially “Top-K routing,” involve a discrete, hard selection process. A common function is KeepTopK(v, k), which selects the top k scoring elements from a vector v and sets others to $-\infty$ or $0$."><meta property="og:url" content="/posts/mixture-of-experts-moe-models-challenges-solutions-in-practice/"><meta property="og:site_name" content="Eric X. Liu's Personal Page"><meta property="og:title" content="Mixture-of-Experts (MoE) Models Challenges & Solutions in Practice"><meta property="og:description" content="Mixture-of-Experts (MoEs) are neural network architectures that allow different parts of the model (called “experts”) to specialize in different types of inputs. A “gating network” or “router” learns to dispatch each input (or “token”) to a subset of these experts. While powerful for scaling models, MoEs introduce several practical challenges.
+1. Challenge: Non-Differentiability of Routing Functions Link to heading The Problem: Many routing mechanisms, especially “Top-K routing,” involve a discrete, hard selection process. A common function is KeepTopK(v, k), which selects the top k scoring elements from a vector v and sets others to $-\infty$ or $0$."><meta property="og:locale" content="en"><meta property="og:type" content="article"><meta property="article:section" content="posts"><meta property="article:published_time" content="2025-07-02T00:00:00+00:00"><meta property="article:modified_time" content="2025-08-03T06:02:48+00:00"><link rel=canonical href=/posts/mixture-of-experts-moe-models-challenges-solutions-in-practice/><link rel=preload href=/fonts/fa-brands-400.woff2 as=font type=font/woff2 crossorigin><link rel=preload href=/fonts/fa-regular-400.woff2 as=font type=font/woff2 crossorigin><link rel=preload href=/fonts/fa-solid-900.woff2 as=font type=font/woff2 crossorigin><link rel=stylesheet href=/css/coder.min.6445a802b9389c9660e1b07b724dcf5718b1065ed2d71b4eeaf981cc7cc5fc46.css integrity="sha256-ZEWoArk4nJZg4bB7ck3PVxixBl7S1xtO6vmBzHzF/EY=" crossorigin=anonymous media=screen><link rel=stylesheet href=/css/coder-dark.min.a00e6364bacbc8266ad1cc81230774a1397198f8cfb7bcba29b7d6fcb54ce57f.css integrity="sha256-oA5jZLrLyCZq0cyBIwd0oTlxmPjPt7y6KbfW/LVM5X8=" crossorigin=anonymous media=screen><link rel=icon type=image/svg+xml href=/images/favicon.svg sizes=any><link rel=icon type=image/png href=/images/favicon-32x32.png sizes=32x32><link rel=icon type=image/png href=/images/favicon-16x16.png sizes=16x16><link rel=apple-touch-icon href=/images/apple-touch-icon.png><link rel=apple-touch-icon sizes=180x180 href=/images/apple-touch-icon.png><link rel=manifest href=/site.webmanifest><link rel=mask-icon href=/images/safari-pinned-tab.svg color=#5bbad5></head><body class="preload-transitions colorscheme-auto"><div class=float-container><a id=dark-mode-toggle class=colorscheme-toggle><i class="fa-solid fa-adjust fa-fw" aria-hidden=true></i></a></div><main class=wrapper><nav class=navigation><section class=container><a class=navigation-title href=/>Eric X. Liu's Personal Page
+</a><input type=checkbox id=menu-toggle>
+<label class="menu-button float-right" for=menu-toggle><i class="fa-solid fa-bars fa-fw" aria-hidden=true></i></label><ul class=navigation-list><li class=navigation-item><a class=navigation-link href=/posts/>Posts</a></li><li class=navigation-item><a class=navigation-link href=https://chat.ericxliu.me>Chat</a></li><li class=navigation-item><a class=navigation-link href=https://git.ericxliu.me/user/oauth2/Authenitk>Git</a></li><li class=navigation-item><a class=navigation-link href=https://coder.ericxliu.me/api/v2/users/oidc/callback>Coder</a></li><li class=navigation-item><a class=navigation-link href=/>|</a></li><li class=navigation-item><a class=navigation-link href=https://sso.ericxliu.me>Sign in</a></li></ul></section></nav><div class=content><section class="container post"><article><header><div class=post-title><h1 class=title><a class=title-link href=/posts/mixture-of-experts-moe-models-challenges-solutions-in-practice/>Mixture-of-Experts (MoE) Models Challenges & Solutions in Practice</a></h1></div><div class=post-meta><div class=date><span class=posted-on><i class="fa-solid fa-calendar" aria-hidden=true></i>
+<time datetime=2025-07-02T00:00:00Z>July 2, 2025
+</time></span><span class=reading-time><i class="fa-solid fa-clock" aria-hidden=true></i>
+7-minute read</span></div></div></header><div class=post-content><p>Mixture-of-Experts (MoEs) are neural network architectures that allow different parts of the model (called &ldquo;experts&rdquo;) to specialize in different types of inputs. A &ldquo;gating network&rdquo; or &ldquo;router&rdquo; learns to dispatch each input (or &ldquo;token&rdquo;) to a subset of these experts. While powerful for scaling models, MoEs introduce several practical challenges.</p><h3 id=1-challenge-non-differentiability-of-routing-functions>1. Challenge: Non-Differentiability of Routing Functions
+<a class=heading-link href=#1-challenge-non-differentiability-of-routing-functions><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><p><strong>The Problem:</strong>
+Many routing mechanisms, especially &ldquo;Top-K routing,&rdquo; involve a discrete, hard selection process. A common function is <code>KeepTopK(v, k)</code>, which selects the top <code>k</code> scoring elements from a vector <code>v</code> and sets others to $-\infty$ or $0$.</p>$$
+KeepTopK(v, k)_i = \begin{cases} v_i & \text{if } v_i \text{ is in the top } k \text{ elements of } v \\ -\infty & \text{otherwise.} \end{cases}
+$$<p>This function is <strong>not differentiable</strong>. Its gradient is zero almost everywhere and undefined at the threshold points, making it impossible to directly train the gating network&rsquo;s parameters (e.g., $W_g$) using standard gradient descent.</p><p><strong>Solutions (Stochastic Approximations):</strong>
+To enable end-to-end training, non-differentiable routing decisions must be approximated with differentiable or stochastic methods.</p><ul><li><p><strong>Stochastic Scoring (e.g., Shazeer et al. 2017):</strong>
+The expert score $H(x)_i = (x \cdot W_g)_i + \text{StandardNormal}() \cdot \text{Softplus}((x \cdot W_{noise})_i)$ introduces Gaussian noise. This makes the scores themselves stochastic, which can be leveraged with other methods.</p></li><li><p><strong>Gumbel-Softmax Trick (or Concrete Distribution):</strong>
+This method allows for differentiable sampling from categorical distributions. Instead of directly picking the top-k, Gumbel noise is added to the scores, and a Softmax (with a temperature parameter) is applied. This provides a continuous, differentiable approximation of a discrete choice, allowing gradients to flow back.</p></li><li><p><strong>REINFORCE (Score Function Estimator):</strong>
+This is a policy gradient method from reinforcement learning. The routing decision is treated as an action, and the gating network&rsquo;s parameters are updated based on the &ldquo;reward&rdquo; (e.g., the model&rsquo;s performance). Gradients are estimated by sampling routing choices and weighting them by their outcomes.</p></li><li><p><strong>Straight-Through Estimator (STE):</strong>
+A simpler approximation where, during the backward pass, gradients are treated as if the non-differentiable operation was an identity function or a simple smooth function.</p></li><li><p><strong>Softmax after TopK (e.g., Mixtral, DBRX, DeepSeek v3):</strong>
+Instead of <code>Softmax(KeepTopK(...))</code>, some models apply a Softmax <em>only to the scores of the selected TopK experts</em>, and then assign $0$ to the rest. This provides differentiable weights for the selected experts while still enforcing sparsity.</p></li></ul><h3 id=2-challenge-uneven-expert-utilization-balancing-loss>2. Challenge: Uneven Expert Utilization (Balancing Loss)
+<a class=heading-link href=#2-challenge-uneven-expert-utilization-balancing-loss><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><p><strong>The Problem:</strong>
+Left unchecked, the gating network might learn to heavily favor a few experts, leaving others underutilized. This leads to:</p><ul><li><strong>System Inefficiency:</strong> Overloaded experts become bottlenecks, while underutilized experts waste computational resources.</li><li><strong>Suboptimal Learning:</strong> Experts might not specialize effectively if they don&rsquo;t receive diverse data.</li></ul><p><strong>Solution: Heuristic Balancing Losses (e.g., from Switch Transformer, Fedus et al. 2022)</strong>
+An auxiliary loss is added to the total model loss during training to encourage more even expert usage.</p>$$ \text{loss}_{\text{auxiliary}} = \alpha \cdot N \cdot \sum_{i=1}^{N} f_i \cdot P_i $$<p>Where:</p><ul><li>$\alpha$: A hyperparameter controlling the strength of the auxiliary loss.</li><li>$N$: Total number of experts.</li><li>$f_i$: The <strong>fraction of tokens <em>actually dispatched</em> to expert $i$</strong> in the current batch $B$.
+$$ f_i = \frac{1}{T} \sum_{x \in B} \mathbf{1}\{\text{argmax } p(x) = i\} $$
+($p(x)$ here refers to the output of the gating network, which could be $s_{i,t}$ in the DeepSeek/classic router. The $\text{argmax}$ means it counts hard assignments to expert $i$.)</li><li>$P_i$: The <strong>fraction of the router <em>probability mass</em> allocated to expert $i$</strong> in the current batch $B$.
+$$ P_i = \frac{1}{T} \sum_{x \in B} p_i(x) $$
+($p_i(x)$ is the learned probability (or soft score) from the gating network for token $x$ and expert $i$.)</li></ul><p><strong>How it works:</strong>
+The loss aims to minimize the product $f_i \cdot P_i$ when $f_i$ and $P_i$ are small, effectively pushing them to be larger (closer to $1/N$). If an expert $i$ is overused (high $f_i$ and $P_i$), its term in the sum contributes significantly to the loss. The derivative with respect to $p_i(x)$ reveals that &ldquo;more frequent use = stronger downweighting,&rdquo; meaning the gating network is penalized for sending too much traffic to an already busy expert.</p><p><strong>Relationship to Gating Network:</strong></p><ul><li><strong>$p_i(x)$ (or $s_{i,t}$):</strong> This is the output of the <strong>learned gating network</strong> (e.g., from a linear layer followed by Softmax). The gating network&rsquo;s parameters are updated via gradient descent, influenced by this auxiliary loss.</li><li><strong>$P_i$:</strong> This is <em>calculated</em> from the outputs of the learned gating network for the current batch. It&rsquo;s not a pre-defined value.</li></ul><p><strong>Limitation (&ldquo;Second Best&rdquo; Scenario):</strong>
+Even with this loss, an expert can remain imbalanced if it&rsquo;s consistently the &ldquo;second best&rdquo; option (high $P_i$) but never the <em>absolute top choice</em> that gets counted in $f_i$ (especially if $K=1$). This is because $f_i$ strictly counts hard assignments based on <code>argmax</code>. This limitation highlights why &ldquo;soft&rdquo; routing or &ldquo;softmax after TopK&rdquo; approaches can be more effective for truly even distribution.</p><h3 id=3-challenge-overfitting-during-fine-tuning>3. Challenge: Overfitting during Fine-tuning
+<a class=heading-link href=#3-challenge-overfitting-during-fine-tuning><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><p><strong>The Problem:</strong>
+Sparse MoE models, despite only activating a few experts per token, possess a very large total number of parameters. When fine-tuning these models on <strong>smaller datasets</strong>, they are highly prone to <strong>overfitting</strong>. The model&rsquo;s vast capacity allows it to memorize the limited fine-tuning data, leading to poor generalization performance on unseen validation data. This is evident when training loss continues to decrease, but validation loss stagnates or increases.</p><p><strong>Solutions:</strong></p><ul><li><p><strong>Zoph et al. Solution – Fine-tune non-MoE MLPs:</strong></p><ul><li>This strategy involves freezing a portion of the MoE model&rsquo;s parameters during fine-tuning, specifically the large expert weights.</li><li>Instead, only the &ldquo;non-MoE&rdquo; parameters (e.g., attention layers, adapter layers, or the gating network itself) are updated.</li><li>This reduces the effective number of trainable parameters during fine-tuning, thereby mitigating the risk of overfitting on small datasets. It assumes the experts are already well-pre-trained for general tasks.</li></ul></li><li><p><strong>DeepSeek Solution – Use Lots of Data (1.4M SFT):</strong></p><ul><li>This approach tackles the problem by providing the model with a very large and diverse dataset for Supervised Fine-Tuning (SFT).</li><li>With abundant data (e.g., 1.4 million examples covering a wide range of tasks and languages), the model&rsquo;s large capacity can be effectively utilized for specialized learning rather than memorization. The diversity and volume of data prevent individual experts from overfitting to specific examples.</li></ul></li></ul><p><strong>Conclusion:</strong>
+MoE models offer significant advantages in terms of model capacity and computational efficiency, but their unique sparse activation pattern introduces challenges in training and fine-tuning. Overcoming non-differentiability in routing and ensuring balanced expert utilization are crucial for effective pre-training. During fine-tuning, managing the model&rsquo;s vast parameter count to prevent overfitting on smaller datasets requires either strategic parameter freezing or access to very large and diverse fine-tuning data.
+The <strong>Top-K routing</strong> mechanism, as illustrated in the provided image, is a core component in many modern Mixture-of-Experts (MoE) models. It involves selecting a fixed number (<code>K</code>) of experts for each input based on relevance scores.</p><hr><p><strong>Traditional Top-K (Deterministic Selection):</strong></p><ul><li><strong>How it works:</strong><ol><li>Calculate relevance scores (<code>s_{i,t}</code>) for each expert <code>i</code> and input <code>t</code>.</li><li>Identify the <code>K</code> experts with the highest scores.</li><li>Experts <em>within</em> the Top-K are assigned their scores (<code>g_{i,t} = s_{i,t}</code>).</li><li>Experts <em>outside</em> the Top-K are assigned a score of <code>0</code> (<code>g_{i,t} = 0</code>).</li><li>The output is a weighted sum of the selected experts&rsquo; outputs.</li></ol></li><li><strong>Pros:</strong> Predictable, deterministic, selects the &ldquo;best&rdquo; experts based on current scores.</li><li><strong>Cons:</strong> Can lead to expert imbalance, where a few popular experts are always chosen, starving others of training.</li></ul><p><strong>Alternative: Sampling from Softmax (Probabilistic Selection):</strong></p><ul><li><strong>How it works:</strong><ol><li>Calculate relevance scores (<code>s_{i,t}</code>) which are treated as probabilities (after softmax).</li><li><strong>Randomly sample</strong> <code>K</code> unique expert indices from the distribution defined by these probabilities.</li><li>Selected experts contribute; unselected experts do not.</li></ol></li><li><strong>Why it&rsquo;s suggested:</strong><ul><li><strong>Load Balancing:</strong> Prevents expert collapse by ensuring all experts get a chance to be selected, even those with slightly lower scores. This promotes more even training across the entire expert pool.</li><li><strong>Diversity & Exploration:</strong> Introduces randomness, potentially leading to better generalization and robustness by exploring different expert combinations.</li></ul></li><li><strong>Pros:</strong> Better load balancing, prevents expert starvation, encourages exploration.</li><li><strong>Cons:</strong> Stochastic (non-deterministic routing), can make debugging harder, might not pick the absolute &ldquo;best&rdquo; expert in a single instance (but better for long-term training).</li></ul><p><strong>Key Takeaway:</strong> While deterministic Top-K is simpler and directly picks the &ldquo;highest-scoring&rdquo; experts, sampling from the softmax offers a more robust training dynamic by ensuring that all experts receive training data, thereby preventing some experts from becoming unused (&ldquo;dead experts&rdquo;).</p><hr></div><footer><div id=disqus_thread></div><script>window.disqus_config=function(){},function(){if(["localhost","127.0.0.1"].indexOf(window.location.hostname)!=-1){document.getElementById("disqus_thread").innerHTML="Disqus comments not available by default when the website is previewed locally.";return}var t=document,e=t.createElement("script");e.async=!0,e.src="//ericxliu-me.disqus.com/embed.js",e.setAttribute("data-timestamp",+new Date),(t.head||t.body).appendChild(e)}(),document.addEventListener("themeChanged",function(){document.readyState=="complete"&&DISQUS.reset({reload:!0,config:disqus_config})})</script></footer></article><link rel=stylesheet href=https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/katex.min.css integrity=sha384-vKruj+a13U8yHIkAyGgK1J3ArTLzrFGBbBc0tDp4ad/EyewESeXE/Iv67Aj8gKZ0 crossorigin=anonymous><script defer src=https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/katex.min.js integrity=sha384-PwRUT/YqbnEjkZO0zZxNqcxACrXe+j766U2amXcgMg5457rve2Y7I6ZJSm2A0mS4 crossorigin=anonymous></script><script defer src=https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/contrib/auto-render.min.js integrity=sha384-+VBxd3r6XgURycqtZ117nYw44OOcIax56Z4dCRWbxyPt0Koah1uHoK0o4+/RRE05 crossorigin=anonymous onload='renderMathInElement(document.body,{delimiters:[{left:"$$",right:"$$",display:!0},{left:"$",right:"$",display:!1},{left:"\\(",right:"\\)",display:!1},{left:"\\[",right:"\\]",display:!0}]})'></script></section></div><footer class=footer><section class=container>©
+2016 -
+2025
+Eric X. Liu
+<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/69c2890">[69c2890]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script><script defer src=https://static.cloudflareinsights.com/beacon.min.js data-cf-beacon='{"token": "987638e636ce4dbb932d038af74c17d1"}'></script></body></html>
--- a/posts/page/1/index.html
+++ b/posts/page/1/index.html
@@ -0,0 +1 @@
+<!doctype html><html lang=en><head><title>/posts/</title><link rel=canonical href=/posts/><meta name=robots content="noindex"><meta charset=utf-8><meta http-equiv=refresh content="0; url=/posts/"></head></html>
--- a/posts/page/2/index.html
+++ b/posts/page/2/index.html
@@ -0,0 +1,8 @@
+<!doctype html><html lang=en><head><title>Posts · Eric X. Liu's Personal Page</title><meta charset=utf-8><meta name=viewport content="width=device-width,initial-scale=1"><meta name=color-scheme content="light dark"><meta name=author content="Eric X. Liu"><meta name=description content="Eric X. Liu - Software & Performance Engineer at Google. Sharing insights about software engineering, performance optimization, tech industry experiences, mountain biking adventures, Jeep overlanding, and outdoor activities."><meta name=keywords content="software engineer,performance engineering,Google engineer,tech blog,software development,performance optimization,Eric Liu,engineering blog,mountain biking,Jeep enthusiast,overlanding,camping,outdoor adventures"><meta name=twitter:card content="summary"><meta name=twitter:title content="Posts"><meta name=twitter:description content="Eric X. Liu - Software & Performance Engineer at Google. Sharing insights about software engineering, performance optimization, tech industry experiences, mountain biking adventures, Jeep overlanding, and outdoor activities."><meta property="og:url" content="/posts/"><meta property="og:site_name" content="Eric X. Liu's Personal Page"><meta property="og:title" content="Posts"><meta property="og:description" content="Eric X. Liu - Software & Performance Engineer at Google. Sharing insights about software engineering, performance optimization, tech industry experiences, mountain biking adventures, Jeep overlanding, and outdoor activities."><meta property="og:locale" content="en"><meta property="og:type" content="website"><link rel=canonical href=/posts/><link rel=preload href=/fonts/fa-brands-400.woff2 as=font type=font/woff2 crossorigin><link rel=preload href=/fonts/fa-regular-400.woff2 as=font type=font/woff2 crossorigin><link rel=preload href=/fonts/fa-solid-900.woff2 as=font type=font/woff2 crossorigin><link rel=stylesheet href=/css/coder.min.6445a802b9389c9660e1b07b724dcf5718b1065ed2d71b4eeaf981cc7cc5fc46.css integrity="sha256-ZEWoArk4nJZg4bB7ck3PVxixBl7S1xtO6vmBzHzF/EY=" crossorigin=anonymous media=screen><link rel=stylesheet href=/css/coder-dark.min.a00e6364bacbc8266ad1cc81230774a1397198f8cfb7bcba29b7d6fcb54ce57f.css integrity="sha256-oA5jZLrLyCZq0cyBIwd0oTlxmPjPt7y6KbfW/LVM5X8=" crossorigin=anonymous media=screen><link rel=icon type=image/svg+xml href=/images/favicon.svg sizes=any><link rel=icon type=image/png href=/images/favicon-32x32.png sizes=32x32><link rel=icon type=image/png href=/images/favicon-16x16.png sizes=16x16><link rel=apple-touch-icon href=/images/apple-touch-icon.png><link rel=apple-touch-icon sizes=180x180 href=/images/apple-touch-icon.png><link rel=manifest href=/site.webmanifest><link rel=mask-icon href=/images/safari-pinned-tab.svg color=#5bbad5><link rel=alternate type=application/rss+xml href=/posts/index.xml title="Eric X. Liu's Personal Page"></head><body class="preload-transitions colorscheme-auto"><div class=float-container><a id=dark-mode-toggle class=colorscheme-toggle><i class="fa-solid fa-adjust fa-fw" aria-hidden=true></i></a></div><main class=wrapper><nav class=navigation><section class=container><a class=navigation-title href=/>Eric X. Liu's Personal Page
+</a><input type=checkbox id=menu-toggle>
+<label class="menu-button float-right" for=menu-toggle><i class="fa-solid fa-bars fa-fw" aria-hidden=true></i></label><ul class=navigation-list><li class=navigation-item><a class=navigation-link href=/posts/>Posts</a></li><li class=navigation-item><a class=navigation-link href=https://chat.ericxliu.me>Chat</a></li><li class=navigation-item><a class=navigation-link href=https://git.ericxliu.me/user/oauth2/Authenitk>Git</a></li><li class=navigation-item><a class=navigation-link href=https://coder.ericxliu.me/api/v2/users/oidc/callback>Coder</a></li><li class=navigation-item><a class=navigation-link href=/>|</a></li><li class=navigation-item><a class=navigation-link href=https://sso.ericxliu.me>Sign in</a></li></ul></section></nav><div class=content><section class="container list"><header><h1 class=title><a class=title-link href=/posts/>Posts</a></h1></header><ul><li><span class=date>October 26, 2020</span>
+<a class=title href=/posts/useful/>Some useful files</a></li></ul><ul class=pagination><li><a href=/posts/>&#171;</a></li><li class=hidden><a href=/posts/>&#8249;</a></li><li><a href=/posts/>1</a></li><li>2</li></ul></section></div><footer class=footer><section class=container>©
+2016 -
+2025
+Eric X. Liu
+<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/69c2890">[69c2890]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script><script defer src=https://static.cloudflareinsights.com/beacon.min.js data-cf-beacon='{"token": "987638e636ce4dbb932d038af74c17d1"}'></script></body></html>
--- a/posts/ppo-for-language-models/index.html
+++ b/posts/ppo-for-language-models/index.html
@@ -0,0 +1,26 @@
+<!doctype html><html lang=en><head><title>A Deep Dive into PPO for Language Models · Eric X. Liu's Personal Page</title><meta charset=utf-8><meta name=viewport content="width=device-width,initial-scale=1"><meta name=color-scheme content="light dark"><meta name=author content="Eric X. Liu"><meta name=description content="Large Language Models (LLMs) have demonstrated astonishing capabilities, but out-of-the-box, they are simply powerful text predictors. They don&rsquo;t inherently understand what makes a response helpful, harmless, or aligned with human values. The technique that has proven most effective at bridging this gap is Reinforcement Learning from Human Feedback (RLHF), and at its heart lies a powerful algorithm: Proximal Policy Optimization (PPO).
+You may have seen diagrams like the one below, which outlines the RLHF training process. It can look intimidating, with a web of interconnected models, losses, and data flows."><meta name=keywords content="software engineer,performance engineering,Google engineer,tech blog,software development,performance optimization,Eric Liu,engineering blog,mountain biking,Jeep enthusiast,overlanding,camping,outdoor adventures"><meta name=twitter:card content="summary"><meta name=twitter:title content="A Deep Dive into PPO for Language Models"><meta name=twitter:description content="Large Language Models (LLMs) have demonstrated astonishing capabilities, but out-of-the-box, they are simply powerful text predictors. They don’t inherently understand what makes a response helpful, harmless, or aligned with human values. The technique that has proven most effective at bridging this gap is Reinforcement Learning from Human Feedback (RLHF), and at its heart lies a powerful algorithm: Proximal Policy Optimization (PPO).
+You may have seen diagrams like the one below, which outlines the RLHF training process. It can look intimidating, with a web of interconnected models, losses, and data flows."><meta property="og:url" content="/posts/ppo-for-language-models/"><meta property="og:site_name" content="Eric X. Liu's Personal Page"><meta property="og:title" content="A Deep Dive into PPO for Language Models"><meta property="og:description" content="Large Language Models (LLMs) have demonstrated astonishing capabilities, but out-of-the-box, they are simply powerful text predictors. They don’t inherently understand what makes a response helpful, harmless, or aligned with human values. The technique that has proven most effective at bridging this gap is Reinforcement Learning from Human Feedback (RLHF), and at its heart lies a powerful algorithm: Proximal Policy Optimization (PPO).
+You may have seen diagrams like the one below, which outlines the RLHF training process. It can look intimidating, with a web of interconnected models, losses, and data flows."><meta property="og:locale" content="en"><meta property="og:type" content="article"><meta property="article:section" content="posts"><meta property="article:published_time" content="2025-08-02T00:00:00+00:00"><meta property="article:modified_time" content="2025-08-20T06:04:36+00:00"><link rel=canonical href=/posts/ppo-for-language-models/><link rel=preload href=/fonts/fa-brands-400.woff2 as=font type=font/woff2 crossorigin><link rel=preload href=/fonts/fa-regular-400.woff2 as=font type=font/woff2 crossorigin><link rel=preload href=/fonts/fa-solid-900.woff2 as=font type=font/woff2 crossorigin><link rel=stylesheet href=/css/coder.min.6445a802b9389c9660e1b07b724dcf5718b1065ed2d71b4eeaf981cc7cc5fc46.css integrity="sha256-ZEWoArk4nJZg4bB7ck3PVxixBl7S1xtO6vmBzHzF/EY=" crossorigin=anonymous media=screen><link rel=stylesheet href=/css/coder-dark.min.a00e6364bacbc8266ad1cc81230774a1397198f8cfb7bcba29b7d6fcb54ce57f.css integrity="sha256-oA5jZLrLyCZq0cyBIwd0oTlxmPjPt7y6KbfW/LVM5X8=" crossorigin=anonymous media=screen><link rel=icon type=image/svg+xml href=/images/favicon.svg sizes=any><link rel=icon type=image/png href=/images/favicon-32x32.png sizes=32x32><link rel=icon type=image/png href=/images/favicon-16x16.png sizes=16x16><link rel=apple-touch-icon href=/images/apple-touch-icon.png><link rel=apple-touch-icon sizes=180x180 href=/images/apple-touch-icon.png><link rel=manifest href=/site.webmanifest><link rel=mask-icon href=/images/safari-pinned-tab.svg color=#5bbad5></head><body class="preload-transitions colorscheme-auto"><div class=float-container><a id=dark-mode-toggle class=colorscheme-toggle><i class="fa-solid fa-adjust fa-fw" aria-hidden=true></i></a></div><main class=wrapper><nav class=navigation><section class=container><a class=navigation-title href=/>Eric X. Liu's Personal Page
+</a><input type=checkbox id=menu-toggle>
+<label class="menu-button float-right" for=menu-toggle><i class="fa-solid fa-bars fa-fw" aria-hidden=true></i></label><ul class=navigation-list><li class=navigation-item><a class=navigation-link href=/posts/>Posts</a></li><li class=navigation-item><a class=navigation-link href=https://chat.ericxliu.me>Chat</a></li><li class=navigation-item><a class=navigation-link href=https://git.ericxliu.me/user/oauth2/Authenitk>Git</a></li><li class=navigation-item><a class=navigation-link href=https://coder.ericxliu.me/api/v2/users/oidc/callback>Coder</a></li><li class=navigation-item><a class=navigation-link href=/>|</a></li><li class=navigation-item><a class=navigation-link href=https://sso.ericxliu.me>Sign in</a></li></ul></section></nav><div class=content><section class="container post"><article><header><div class=post-title><h1 class=title><a class=title-link href=/posts/ppo-for-language-models/>A Deep Dive into PPO for Language Models</a></h1></div><div class=post-meta><div class=date><span class=posted-on><i class="fa-solid fa-calendar" aria-hidden=true></i>
+<time datetime=2025-08-02T00:00:00Z>August 2, 2025
+</time></span><span class=reading-time><i class="fa-solid fa-clock" aria-hidden=true></i>
+7-minute read</span></div></div></header><div class=post-content><p>Large Language Models (LLMs) have demonstrated astonishing capabilities, but out-of-the-box, they are simply powerful text predictors. They don&rsquo;t inherently understand what makes a response helpful, harmless, or aligned with human values. The technique that has proven most effective at bridging this gap is Reinforcement Learning from Human Feedback (RLHF), and at its heart lies a powerful algorithm: Proximal Policy Optimization (PPO).</p><p>You may have seen diagrams like the one below, which outlines the RLHF training process. It can look intimidating, with a web of interconnected models, losses, and data flows.</p><p><img src=/images/ppo-for-language-models/.png alt></p><p>This post will decode that diagram, piece by piece. We&rsquo;ll explore the &ldquo;why&rdquo; behind each component, moving from high-level concepts to the deep technical reasoning that makes this process work.</p><h3 id=translating-rl-to-a-conversation>Translating RL to a Conversation
+<a class=heading-link href=#translating-rl-to-a-conversation><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><p>The first step is to understand how the traditional language of reinforcement learning maps to the world of text generation.</p><ul><li><strong>State (<code>s_t</code>)</strong>: In a chat setting, the &ldquo;state&rdquo; is the context of the conversation so far. It&rsquo;s the initial prompt (<code>x</code>) plus all the text the model has generated up to the current moment (<code>y₁, ..., y_{t-1}</code>).</li><li><strong>Action (<code>a_t</code>)</strong>: The &ldquo;action&rdquo; is the model&rsquo;s decision at each step. For an LLM, this means generating the very next token (<code>y_t</code>). A full response is a sequence of these actions.blob:https://aistudio.google.com/872e746f-88c1-40ec-8e45-fa0efce97299</li><li><strong>Reward (<code>r</code>)</strong>: The &ldquo;reward&rdquo; is a numeric score that tells the model how good its full response (<code>y</code>) was. This score comes from a separate <strong>Reward Model</strong>, which has been trained on a large dataset of human preference comparisons (e.g., humans rating which of two responses is better). This reward is often only awarded at the end of the entire generated sequence.</li></ul><p>Let&rsquo;s make this concrete. If a user provides the prompt <strong>(x)</strong>: <em>&ldquo;The best thing about AI is&rdquo;</em>, and the model generates the response <strong>(y)</strong>: <em>&ldquo;its potential to solve problems.&rdquo;</em>, here is how it&rsquo;s broken down for training:</p><ul><li><strong>State 1</strong>: &ldquo;The best thing about AI is&rdquo;<ul><li><strong>Action 1</strong>: &ldquo;its&rdquo;</li></ul></li><li><strong>State 2</strong>: &ldquo;The best thing about AI is its&rdquo;<ul><li><strong>Action 2</strong>: " potential"</li></ul></li><li><strong>State 3</strong>: &ldquo;The best thing about AI is its potential&rdquo;<ul><li><strong>Action 3</strong>: " to"</li></ul></li><li>&mldr;and so on for every generated token.</li></ul><p>This breakdown transforms a single prompt-response pair into a rich trajectory of state-action pairs, which becomes the raw data for our learning algorithm.</p><h3 id=the-cast-of-models-an-actor-critic-ensemble>The Cast of Models: An Actor-Critic Ensemble
+<a class=heading-link href=#the-cast-of-models-an-actor-critic-ensemble><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><p>The PPO process doesn&rsquo;t rely on a single model but an ensemble where each member has a distinct role.</p><ol><li><strong>The Actor (Policy LM)</strong>: This is the star of the show—the LLM we are actively fine-tuning. Its role is to take a state (the current text) and decide on an action (the next token). We refer to its decision-making process as its &ldquo;policy&rdquo; (<code>π</code>).</li><li><strong>The Critic (Value Model)</strong>: This is the Actor&rsquo;s coach. The Critic doesn&rsquo;t generate text. Instead, it observes a state and estimates the <em>potential future reward</em> the Actor is likely to receive from that point onward. This estimate is called the &ldquo;value&rdquo; (<code>V(s_t)</code>). The Critic&rsquo;s feedback helps the Actor understand whether it&rsquo;s in a promising or a dead-end situation, which is a much more immediate learning signal than waiting for the final reward.</li><li><strong>The Reward Model</strong>: This is the ultimate judge. As mentioned, it&rsquo;s a separate model trained on human preference data that provides the final score for a complete generation. Its judgment is treated as the ground truth for training both the Actor and the Critic.</li></ol><h3 id=the-challenge-of-credit-assignment-generalized-advantage-estimation-gae>The Challenge of Credit Assignment: Generalized Advantage Estimation (GAE)
+<a class=heading-link href=#the-challenge-of-credit-assignment-generalized-advantage-estimation-gae><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><p>A key problem in RL is assigning credit. If a 20-token response gets a high reward, was it because of the first token, the last one, or all of them? The Critic helps solve this. By comparing the reward at each step with the Critic&rsquo;s value estimate, we can calculate the <strong>Advantage (<code>Â</code>)</strong>.</p><p>A simple advantage calculation might be: <code>Advantage = reward + Value_of_next_state - Value_of_current_state</code>.</p><p>However, this can be noisy. PPO uses a more sophisticated technique called <strong>Generalized Advantage Estimation (GAE)</strong>. The formula looks complex, but the idea is intuitive:</p><p><code>Â(s_t, a_t) = Σ(γλ)^l * δ_{t+l}</code>
+where <code>δ_t = r_t + γV(s_{t+1}) - V(s_t)</code></p><ul><li><strong>γ (gamma)</strong> is a discount factor (e.g., 0.99), which values immediate rewards slightly more than distant ones.</li><li><strong>λ (lambda)</strong> is a smoothing parameter that balances the trade-off between bias and variance. It creates a weighted average of advantages over multiple future time steps.</li></ul><p>In essence, GAE provides a more stable and accurate estimate of how much better a specific action was compared to the policy&rsquo;s average behavior in that state.</p><h3 id=the-heart-of-ppo-the-quest-for-stable-updates>The Heart of PPO: The Quest for Stable Updates
+<a class=heading-link href=#the-heart-of-ppo-the-quest-for-stable-updates><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><p>Now we arrive at the core innovation of PPO. We want to update our Actor model to take actions with higher advantages. The naive way to do this is to re-weight our training objective by an <strong>importance sampling ratio</strong>: <code>(π_new / π_old)</code>. This corrects for the fact that the data we are learning from was generated by a slightly older version of our policy.</p><p>However, this ratio is incredibly dangerous. If the new policy becomes very different from the old one, the ratio can explode, leading to massive, unstable gradient updates that destroy the model.</p><p>PPO solves this with its signature <strong>Clipped Surrogate Objective</strong>. The PPO loss function is:</p><p><code>L_CLIP(θ) = Ê_t [ min( r_t(θ)Â_t, clip(r_t(θ), 1 - ε, 1 + ε)Â_t ) ]</code></p><p>Let&rsquo;s translate this from math to English:</p><ul><li><code>r_t(θ)</code> is the probability ratio <code>π_new(a_t|s_t) / π_old(a_t|s_t)</code>.</li><li>The goal is to increase the objective by an amount proportional to the advantage <code>Â_t</code>.</li><li><strong>The <code>clip</code> function is the crucial safeguard.</strong> It forbids the probability ratio from moving outside a small window (e.g., <code>[0.8, 1.2]</code>).</li></ul><p>This means the algorithm says: &ldquo;Let&rsquo;s update our policy to favor this good action. But if the required update would change the policy too drastically from the old one, we&rsquo;ll &lsquo;clip&rsquo; the update to a more modest size.&rdquo; This creates a &ldquo;trust region,&rdquo; ensuring stable, incremental improvements.</p><h3 id=avoiding-amnesia-the-pretraining-loss>Avoiding Amnesia: The Pretraining Loss
+<a class=heading-link href=#avoiding-amnesia-the-pretraining-loss><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><p>There&rsquo;s one final problem. If we only optimize for the PPO loss, the model might learn to &ldquo;hack&rdquo; the reward model by generating repetitive or nonsensical text that gets a high score. In doing so, it could suffer from <strong>catastrophic forgetting</strong>, losing its fundamental grasp of grammar and facts.</p><p>To prevent this, we introduce a second loss term. As seen in the diagram, we mix in data from the original <strong>Pretraining Data</strong> (or the dataset used for Supervised Fine-Tuning). We calculate a standard next-token prediction loss (<code>LM Loss</code>) on this high-quality data.</p><p>The final loss for the Actor is a combination of both objectives:</p><p><strong>Total Loss = Loss_PPO + <code>λ_ptx</code> * Loss_LM</strong></p><p>This brilliantly balances two goals:</p><ol><li>The <code>Loss_PPO</code> pushes the model towards behaviors that align with human preferences.</li><li>The <code>Loss_LM</code> acts as a regularizer, pulling the model back towards its core language capabilities and preventing it from drifting into gibberish.</li></ol><h3 id=the-full-training-loop>The Full Training Loop
+<a class=heading-link href=#the-full-training-loop><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><p>Now, we can assemble the entire process into a clear, iterative loop:</p><ol><li><strong>Collect</strong>: The current Actor policy <code>π_k</code> generates responses to a batch of prompts. These experiences—<code>(state, action, probability, reward, value)</code>—are stored in an <strong>Experience Buffer</strong>.</li><li><strong>Calculate</strong>: Once the buffer is full, we use the collected data to compute the advantage estimates <code>Â_t</code> for every single token-generation step.</li><li><strong>Optimize</strong>: For a few epochs, we repeatedly sample mini-batches from the buffer and update the Actor and Critic models. The Actor is updated using the combined <code>PPO-clip Loss</code> and <code>LM Loss</code>. The Critic is updated to improve its value predictions.</li><li><strong>Flush and Repeat</strong>: After the optimization phase, the entire experience buffer is discarded. The data is now &ldquo;stale&rdquo; because our policy has changed. The newly updated policy <code>π_{k+1}</code> becomes the new Actor, and we return to step 1 to collect fresh data.</li></ol><p>This cycle of collection and optimization allows the language model to gradually and safely steer its behavior towards human-defined goals, creating the helpful and aligned AI assistants we interact with today.</p><hr><p><strong>References:</strong></p><ol><li>Schulman, J., Wolski, F., Dhariwal, P., Radford, A., & Klimov, O. (2017). <em>Proximal Policy Optimization Algorithms</em>. arXiv preprint arXiv:1707.06347.</li><li>Schulman, J., Moritz, P., Levine, S., Jordan, M., & Abbeel, P. (2015). <em>High-Dimensional Continuous Control Using Generalized Advantage Estimation</em>. arXiv preprint arXiv:1506.02438.</li><li>Ouyang, L., et al. (2022). <em>Training language models to follow instructions with human feedback</em>. Advances in Neural Information Processing Systems 35.</li></ol></div><footer><div id=disqus_thread></div><script>window.disqus_config=function(){},function(){if(["localhost","127.0.0.1"].indexOf(window.location.hostname)!=-1){document.getElementById("disqus_thread").innerHTML="Disqus comments not available by default when the website is previewed locally.";return}var t=document,e=t.createElement("script");e.async=!0,e.src="//ericxliu-me.disqus.com/embed.js",e.setAttribute("data-timestamp",+new Date),(t.head||t.body).appendChild(e)}(),document.addEventListener("themeChanged",function(){document.readyState=="complete"&&DISQUS.reset({reload:!0,config:disqus_config})})</script></footer></article><link rel=stylesheet href=https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/katex.min.css integrity=sha384-vKruj+a13U8yHIkAyGgK1J3ArTLzrFGBbBc0tDp4ad/EyewESeXE/Iv67Aj8gKZ0 crossorigin=anonymous><script defer src=https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/katex.min.js integrity=sha384-PwRUT/YqbnEjkZO0zZxNqcxACrXe+j766U2amXcgMg5457rve2Y7I6ZJSm2A0mS4 crossorigin=anonymous></script><script defer src=https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/contrib/auto-render.min.js integrity=sha384-+VBxd3r6XgURycqtZ117nYw44OOcIax56Z4dCRWbxyPt0Koah1uHoK0o4+/RRE05 crossorigin=anonymous onload='renderMathInElement(document.body,{delimiters:[{left:"$$",right:"$$",display:!0},{left:"$",right:"$",display:!1},{left:"\\(",right:"\\)",display:!1},{left:"\\[",right:"\\]",display:!0}]})'></script></section></div><footer class=footer><section class=container>©
+2016 -
+2025
+Eric X. Liu
+<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/69c2890">[69c2890]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script><script defer src=https://static.cloudflareinsights.com/beacon.min.js data-cf-beacon='{"token": "987638e636ce4dbb932d038af74c17d1"}'></script></body></html>
--- a/posts/quantization-in-llms/index.html
+++ b/posts/quantization-in-llms/index.html
--- a/posts/secure-boot-dkms-and-mok-on-proxmox-debian/index.html
+++ b/posts/secure-boot-dkms-and-mok-on-proxmox-debian/index.html
@@ -0,0 +1,62 @@
+<!doctype html><html lang=en><head><title>Fixing GPU Operator Pods Stuck in Init: Secure Boot, DKMS, and MOK on Proxmox + Debian · Eric X. Liu's Personal Page</title><meta charset=utf-8><meta name=viewport content="width=device-width,initial-scale=1"><meta name=color-scheme content="light dark"><meta name=author content="Eric X. Liu"><meta name=description content="I hit an issue where all GPU Operator pods on one node were stuck in Init after migrating from Legacy BIOS to UEFI. The common error was NVIDIA components waiting for “toolkit-ready,” while the toolkit init container looped with:
+
+nvidia-smi failed to communicate with the NVIDIA driver
+modprobe nvidia → “Key was rejected by service”
+
+That message is the tell: Secure Boot is enabled and the kernel refuses to load modules not signed by a trusted key."><meta name=keywords content="software engineer,performance engineering,Google engineer,tech blog,software development,performance optimization,Eric Liu,engineering blog,mountain biking,Jeep enthusiast,overlanding,camping,outdoor adventures"><meta name=twitter:card content="summary"><meta name=twitter:title content="Fixing GPU Operator Pods Stuck in Init: Secure Boot, DKMS, and MOK on Proxmox + Debian"><meta name=twitter:description content="I hit an issue where all GPU Operator pods on one node were stuck in Init after migrating from Legacy BIOS to UEFI. The common error was NVIDIA components waiting for “toolkit-ready,” while the toolkit init container looped with:
+nvidia-smi failed to communicate with the NVIDIA driver modprobe nvidia → “Key was rejected by service” That message is the tell: Secure Boot is enabled and the kernel refuses to load modules not signed by a trusted key."><meta property="og:url" content="/posts/secure-boot-dkms-and-mok-on-proxmox-debian/"><meta property="og:site_name" content="Eric X. Liu's Personal Page"><meta property="og:title" content="Fixing GPU Operator Pods Stuck in Init: Secure Boot, DKMS, and MOK on Proxmox + Debian"><meta property="og:description" content="I hit an issue where all GPU Operator pods on one node were stuck in Init after migrating from Legacy BIOS to UEFI. The common error was NVIDIA components waiting for “toolkit-ready,” while the toolkit init container looped with:
+nvidia-smi failed to communicate with the NVIDIA driver modprobe nvidia → “Key was rejected by service” That message is the tell: Secure Boot is enabled and the kernel refuses to load modules not signed by a trusted key."><meta property="og:locale" content="en"><meta property="og:type" content="article"><meta property="article:section" content="posts"><meta property="article:published_time" content="2025-08-09T00:00:00+00:00"><meta property="article:modified_time" content="2025-08-14T06:50:22+00:00"><link rel=canonical href=/posts/secure-boot-dkms-and-mok-on-proxmox-debian/><link rel=preload href=/fonts/fa-brands-400.woff2 as=font type=font/woff2 crossorigin><link rel=preload href=/fonts/fa-regular-400.woff2 as=font type=font/woff2 crossorigin><link rel=preload href=/fonts/fa-solid-900.woff2 as=font type=font/woff2 crossorigin><link rel=stylesheet href=/css/coder.min.6445a802b9389c9660e1b07b724dcf5718b1065ed2d71b4eeaf981cc7cc5fc46.css integrity="sha256-ZEWoArk4nJZg4bB7ck3PVxixBl7S1xtO6vmBzHzF/EY=" crossorigin=anonymous media=screen><link rel=stylesheet href=/css/coder-dark.min.a00e6364bacbc8266ad1cc81230774a1397198f8cfb7bcba29b7d6fcb54ce57f.css integrity="sha256-oA5jZLrLyCZq0cyBIwd0oTlxmPjPt7y6KbfW/LVM5X8=" crossorigin=anonymous media=screen><link rel=icon type=image/svg+xml href=/images/favicon.svg sizes=any><link rel=icon type=image/png href=/images/favicon-32x32.png sizes=32x32><link rel=icon type=image/png href=/images/favicon-16x16.png sizes=16x16><link rel=apple-touch-icon href=/images/apple-touch-icon.png><link rel=apple-touch-icon sizes=180x180 href=/images/apple-touch-icon.png><link rel=manifest href=/site.webmanifest><link rel=mask-icon href=/images/safari-pinned-tab.svg color=#5bbad5></head><body class="preload-transitions colorscheme-auto"><div class=float-container><a id=dark-mode-toggle class=colorscheme-toggle><i class="fa-solid fa-adjust fa-fw" aria-hidden=true></i></a></div><main class=wrapper><nav class=navigation><section class=container><a class=navigation-title href=/>Eric X. Liu's Personal Page
+</a><input type=checkbox id=menu-toggle>
+<label class="menu-button float-right" for=menu-toggle><i class="fa-solid fa-bars fa-fw" aria-hidden=true></i></label><ul class=navigation-list><li class=navigation-item><a class=navigation-link href=/posts/>Posts</a></li><li class=navigation-item><a class=navigation-link href=https://chat.ericxliu.me>Chat</a></li><li class=navigation-item><a class=navigation-link href=https://git.ericxliu.me/user/oauth2/Authenitk>Git</a></li><li class=navigation-item><a class=navigation-link href=https://coder.ericxliu.me/api/v2/users/oidc/callback>Coder</a></li><li class=navigation-item><a class=navigation-link href=/>|</a></li><li class=navigation-item><a class=navigation-link href=https://sso.ericxliu.me>Sign in</a></li></ul></section></nav><div class=content><section class="container post"><article><header><div class=post-title><h1 class=title><a class=title-link href=/posts/secure-boot-dkms-and-mok-on-proxmox-debian/>Fixing GPU Operator Pods Stuck in Init: Secure Boot, DKMS, and MOK on Proxmox + Debian</a></h1></div><div class=post-meta><div class=date><span class=posted-on><i class="fa-solid fa-calendar" aria-hidden=true></i>
+<time datetime=2025-08-09T00:00:00Z>August 9, 2025
+</time></span><span class=reading-time><i class="fa-solid fa-clock" aria-hidden=true></i>
+3-minute read</span></div></div></header><div class=post-content><p>I hit an issue where all GPU Operator pods on one node were stuck in Init after migrating from Legacy BIOS to UEFI. The common error was NVIDIA components waiting for “toolkit-ready,” while the toolkit init container looped with:</p><ul><li>nvidia-smi failed to communicate with the NVIDIA driver</li><li>modprobe nvidia → “Key was rejected by service”</li></ul><p>That message is the tell: Secure Boot is enabled and the kernel refuses to load modules not signed by a trusted key.</p><h3 id=environment>Environment
+<a class=heading-link href=#environment><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><ul><li>Proxmox VM (QEMU/KVM) 8.4.9</li><li>Debian 12 (bookworm), kernel 6.1</li><li>GPU: NVIDIA Tesla V100 (GV100GL)</li><li>NVIDIA driver installed via Debian packages (nvidia-driver, nvidia-kernel-dkms)</li></ul><h3 id=root-cause>Root Cause
+<a class=heading-link href=#root-cause><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><ul><li>Secure Boot enabled (verified with <code>mokutil --sb-state</code>)</li><li>NVIDIA DKMS modules were built, but the signing key was not trusted by the UEFI shim/firmware</li><li>VM booted via the fallback “UEFI QEMU HARDDISK” path (not shim), so MOK requests didn’t run; no MOK screen</li></ul><h3 id=strategy>Strategy
+<a class=heading-link href=#strategy><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><p>Keep Secure Boot on; get modules trusted. That requires:</p><ol><li>Ensure the VM boots via shim (so MOK can work)</li><li>Make sure DKMS signs modules with a MOK key/cert</li><li>Enroll that MOK into the firmware via shim’s MokManager</li></ol><h3 id=step-1--boot-via-shim-and-persist-efi-variables>Step 1 — Boot via shim and persist EFI variables
+<a class=heading-link href=#step-1--boot-via-shim-and-persist-efi-variables><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><p>In Proxmox (VM stopped):</p><ul><li>BIOS: OVMF (UEFI)</li><li>Add EFI Disk (stores OVMF VARS; required for MOK)</li><li>Machine: q35</li><li>Enable Secure Boot (option shows only with OVMF + EFI Disk)</li></ul><p>Inside Debian:</p><ul><li>Ensure ESP is mounted at <code>/boot/efi</code></li><li>Install signed boot stack:<div class=highlight><pre tabindex=0 style=color:#e6edf3;background-color:#0d1117;-moz-tab-size:4;-o-tab-size:4;tab-size:4><code class=language-bash data-lang=bash><span style=display:flex><span>sudo apt install shim-signed grub-efi-amd64-signed efibootmgr mokutil
+</span></span><span style=display:flex><span>sudo grub-install --target<span style=color:#ff7b72;font-weight:700>=</span>x86_64-efi --efi-directory<span style=color:#ff7b72;font-weight:700>=</span>/boot/efi --bootloader-id<span style=color:#ff7b72;font-weight:700>=</span>debian
+</span></span><span style=display:flex><span>sudo update-grub
+</span></span></code></pre></div></li><li>Create/verify a boot entry that points to shim:<div class=highlight><pre tabindex=0 style=color:#e6edf3;background-color:#0d1117;-moz-tab-size:4;-o-tab-size:4;tab-size:4><code class=language-bash data-lang=bash><span style=display:flex><span>sudo efibootmgr -c -d /dev/sda -p <span style=color:#a5d6ff>15</span> -L <span style=color:#a5d6ff>&#34;debian&#34;</span> -l <span style=color:#a5d6ff>&#39;\EFI\debian\shimx64.efi&#39;</span>
+</span></span><span style=display:flex><span>sudo efibootmgr -o 0002,0001,0000     <span style=color:#8b949e;font-style:italic># make shim (0002) first</span>
+</span></span><span style=display:flex><span>sudo efibootmgr -n <span style=color:#a5d6ff>0002</span>               <span style=color:#8b949e;font-style:italic># BootNext shim for the next reboot</span>
+</span></span></code></pre></div></li></ul><p>Tip: If NVRAM resets or fallback path is used, copy as a fallback:</p><div class=highlight><pre tabindex=0 style=color:#e6edf3;background-color:#0d1117;-moz-tab-size:4;-o-tab-size:4;tab-size:4><code class=language-bash data-lang=bash><span style=display:flex><span>sudo mkdir -p /boot/efi/EFI/BOOT
+</span></span><span style=display:flex><span>sudo cp /boot/efi/EFI/debian/shimx64.efi /boot/efi/EFI/BOOT/BOOTX64.EFI
+</span></span><span style=display:flex><span>sudo cp /boot/efi/EFI/debian/<span style=color:#ff7b72;font-weight:700>{</span>mmx64.efi,grubx64.efi<span style=color:#ff7b72;font-weight:700>}</span> /boot/efi/EFI/BOOT/
+</span></span></code></pre></div><h3 id=step-2--make-dkms-sign-nvidia-modules-with-a-mok>Step 2 — Make DKMS sign NVIDIA modules with a MOK
+<a class=heading-link href=#step-2--make-dkms-sign-nvidia-modules-with-a-mok><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><p>Debian already generated a DKMS key at <code>/var/lib/dkms/mok.key</code>. Create an X.509 cert in DER format:</p><div class=highlight><pre tabindex=0 style=color:#e6edf3;background-color:#0d1117;-moz-tab-size:4;-o-tab-size:4;tab-size:4><code class=language-bash data-lang=bash><span style=display:flex><span>sudo openssl req -new -x509 <span style=color:#79c0ff>\
+</span></span></span><span style=display:flex><span><span style=color:#79c0ff></span>  -key /var/lib/dkms/mok.key <span style=color:#79c0ff>\
+</span></span></span><span style=display:flex><span><span style=color:#79c0ff></span>  -out /var/lib/dkms/mok.der <span style=color:#79c0ff>\
+</span></span></span><span style=display:flex><span><span style=color:#79c0ff></span>  -outform DER <span style=color:#79c0ff>\
+</span></span></span><span style=display:flex><span><span style=color:#79c0ff></span>  -subj <span style=color:#a5d6ff>&#34;/CN=DKMS MOK/&#34;</span> <span style=color:#79c0ff>\
+</span></span></span><span style=display:flex><span><span style=color:#79c0ff></span>  -days <span style=color:#a5d6ff>36500</span>
+</span></span></code></pre></div><p>Enable DKMS signing:</p><div class=highlight><pre tabindex=0 style=color:#e6edf3;background-color:#0d1117;-moz-tab-size:4;-o-tab-size:4;tab-size:4><code class=language-bash data-lang=bash><span style=display:flex><span>sudo sed -i <span style=color:#a5d6ff>&#39;s|^mok_signing_key=.*|mok_signing_key=/var/lib/dkms/mok.key|&#39;</span> /etc/dkms/framework.conf
+</span></span><span style=display:flex><span>sudo sed -i <span style=color:#a5d6ff>&#39;s|^mok_certificate=.*|mok_certificate=/var/lib/dkms/mok.der|&#39;</span> /etc/dkms/framework.conf
+</span></span></code></pre></div><p>Rebuild/install modules (signs them now):</p><div class=highlight><pre tabindex=0 style=color:#e6edf3;background-color:#0d1117;-moz-tab-size:4;-o-tab-size:4;tab-size:4><code class=language-bash data-lang=bash><span style=display:flex><span>sudo dkms build nvidia/<span style=color:#ff7b72>$(</span>modinfo -F version nvidia<span style=color:#ff7b72>)</span> -k <span style=color:#ff7b72>$(</span>uname -r<span style=color:#ff7b72>)</span> --force
+</span></span><span style=display:flex><span>sudo dkms install nvidia/<span style=color:#ff7b72>$(</span>modinfo -F version nvidia<span style=color:#ff7b72>)</span> -k <span style=color:#ff7b72>$(</span>uname -r<span style=color:#ff7b72>)</span> --force
+</span></span></code></pre></div><h3 id=step-3--enroll-the-mok-via-shim-mokmanager>Step 3 — Enroll the MOK via shim (MokManager)
+<a class=heading-link href=#step-3--enroll-the-mok-via-shim-mokmanager><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><p>Queue the cert and set a longer prompt timeout:</p><div class=highlight><pre tabindex=0 style=color:#e6edf3;background-color:#0d1117;-moz-tab-size:4;-o-tab-size:4;tab-size:4><code class=language-bash data-lang=bash><span style=display:flex><span>sudo mokutil --revoke-import
+</span></span><span style=display:flex><span>sudo mokutil --import /var/lib/dkms/mok.der
+</span></span><span style=display:flex><span>sudo mokutil --timeout <span style=color:#a5d6ff>30</span>
+</span></span><span style=display:flex><span>sudo efibootmgr -n <span style=color:#a5d6ff>0002</span>  <span style=color:#8b949e;font-style:italic># ensure next boot goes through shim</span>
+</span></span></code></pre></div><p>Reboot to the VM console (not SSH). In the blue MOK UI:</p><ul><li>Enroll MOK → Continue → Yes → enter password → reboot</li></ul><p>If arrow keys don’t work in Proxmox noVNC:</p><ul><li>Use SPICE (virt-viewer), or</li><li>From the Proxmox host, send keys:<ul><li><code>qm sendkey &lt;VMID> down</code>, <code>qm sendkey &lt;VMID> ret</code>, <code>qm sendkey &lt;VMID> esc</code></li></ul></li></ul><h3 id=verification>Verification
+<a class=heading-link href=#verification><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><div class=highlight><pre tabindex=0 style=color:#e6edf3;background-color:#0d1117;-moz-tab-size:4;-o-tab-size:4;tab-size:4><code class=language-bash data-lang=bash><span style=display:flex><span>sudo mokutil --test-key /var/lib/dkms/mok.der   <span style=color:#8b949e;font-style:italic># “already enrolled”</span>
+</span></span><span style=display:flex><span>sudo modprobe nvidia
+</span></span><span style=display:flex><span>nvidia-smi
+</span></span><span style=display:flex><span>kubectl -n gpu-operator get pods -o wide
+</span></span></code></pre></div><p>Once the module loads, GPU Operator pods on that node leave Init and become Ready.</p><h3 id=key-insights>Key Insights
+<a class=heading-link href=#key-insights><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><ul><li>“Key was rejected by service” during <code>modprobe nvidia</code> means Secure Boot rejected an untrusted module.</li><li>Without shim in the boot path (or without a persistent EFI vars disk), <code>mokutil --import</code> won’t surface a MOK screen.</li><li>DKMS will not sign modules unless configured; set <code>mok_signing_key</code> and <code>mok_certificate</code> in <code>/etc/dkms/framework.conf</code>.</li><li>If you cannot or don’t want to use MOK, the pragmatic dev choice is to disable Secure Boot in OVMF. For production, prefer shim+MOK.</li></ul><h3 id=references>References
+<a class=heading-link href=#references><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><ul><li>Proxmox Secure Boot setup (shim + MOK, EFI vars, DKMS): <a href=https://pve.proxmox.com/wiki/Secure_Boot_Setup#Setup_instructions_for_shim_+_MOK_variant class=external-link target=_blank rel=noopener>Proxmox docs</a></li></ul></div><footer><div id=disqus_thread></div><script>window.disqus_config=function(){},function(){if(["localhost","127.0.0.1"].indexOf(window.location.hostname)!=-1){document.getElementById("disqus_thread").innerHTML="Disqus comments not available by default when the website is previewed locally.";return}var t=document,e=t.createElement("script");e.async=!0,e.src="//ericxliu-me.disqus.com/embed.js",e.setAttribute("data-timestamp",+new Date),(t.head||t.body).appendChild(e)}(),document.addEventListener("themeChanged",function(){document.readyState=="complete"&&DISQUS.reset({reload:!0,config:disqus_config})})</script></footer></article><link rel=stylesheet href=https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/katex.min.css integrity=sha384-vKruj+a13U8yHIkAyGgK1J3ArTLzrFGBbBc0tDp4ad/EyewESeXE/Iv67Aj8gKZ0 crossorigin=anonymous><script defer src=https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/katex.min.js integrity=sha384-PwRUT/YqbnEjkZO0zZxNqcxACrXe+j766U2amXcgMg5457rve2Y7I6ZJSm2A0mS4 crossorigin=anonymous></script><script defer src=https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/contrib/auto-render.min.js integrity=sha384-+VBxd3r6XgURycqtZ117nYw44OOcIax56Z4dCRWbxyPt0Koah1uHoK0o4+/RRE05 crossorigin=anonymous onload='renderMathInElement(document.body,{delimiters:[{left:"$$",right:"$$",display:!0},{left:"$",right:"$",display:!1},{left:"\\(",right:"\\)",display:!1},{left:"\\[",right:"\\]",display:!0}]})'></script></section></div><footer class=footer><section class=container>©
+2016 -
+2025
+Eric X. Liu
+<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/69c2890">[69c2890]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script><script defer src=https://static.cloudflareinsights.com/beacon.min.js data-cf-beacon='{"token": "987638e636ce4dbb932d038af74c17d1"}'></script></body></html>
--- a/posts/supabase-deep-dive/index.html
+++ b/posts/supabase-deep-dive/index.html
@@ -0,0 +1,93 @@
+<!doctype html><html lang=en><head><title>Supabase Deep Dive: It's Not Magic, It's Just Postgres · Eric X. Liu's Personal Page</title><meta charset=utf-8><meta name=viewport content="width=device-width,initial-scale=1"><meta name=color-scheme content="light dark"><meta name=author content="Eric X. Liu"><meta name=description content="In the world of Backend-as-a-Service (BaaS), platforms are often treated as magic boxes. You push data in, you get data out, and you hope the magic inside scales. While this simplicity is powerful, it can obscure the underlying mechanics, leaving developers wondering what&rsquo;s really going on.
+Supabase enters this space with a radically different philosophy: transparency. It provides the convenience of a BaaS, but it’s built on the world&rsquo;s most trusted relational database: PostgreSQL. The &ldquo;magic&rdquo; isn&rsquo;t a proprietary black box; it&rsquo;s a carefully assembled suite of open-source tools that enhance Postgres, not hide it."><meta name=keywords content="software engineer,performance engineering,Google engineer,tech blog,software development,performance optimization,Eric Liu,engineering blog,mountain biking,Jeep enthusiast,overlanding,camping,outdoor adventures"><meta name=twitter:card content="summary"><meta name=twitter:title content="Supabase Deep Dive: It's Not Magic, It's Just Postgres"><meta name=twitter:description content="In the world of Backend-as-a-Service (BaaS), platforms are often treated as magic boxes. You push data in, you get data out, and you hope the magic inside scales. While this simplicity is powerful, it can obscure the underlying mechanics, leaving developers wondering what’s really going on.
+Supabase enters this space with a radically different philosophy: transparency. It provides the convenience of a BaaS, but it’s built on the world’s most trusted relational database: PostgreSQL. The “magic” isn’t a proprietary black box; it’s a carefully assembled suite of open-source tools that enhance Postgres, not hide it."><meta property="og:url" content="/posts/supabase-deep-dive/"><meta property="og:site_name" content="Eric X. Liu's Personal Page"><meta property="og:title" content="Supabase Deep Dive: It's Not Magic, It's Just Postgres"><meta property="og:description" content="In the world of Backend-as-a-Service (BaaS), platforms are often treated as magic boxes. You push data in, you get data out, and you hope the magic inside scales. While this simplicity is powerful, it can obscure the underlying mechanics, leaving developers wondering what’s really going on.
+Supabase enters this space with a radically different philosophy: transparency. It provides the convenience of a BaaS, but it’s built on the world’s most trusted relational database: PostgreSQL. The “magic” isn’t a proprietary black box; it’s a carefully assembled suite of open-source tools that enhance Postgres, not hide it."><meta property="og:locale" content="en"><meta property="og:type" content="article"><meta property="article:section" content="posts"><meta property="article:published_time" content="2025-08-03T00:00:00+00:00"><meta property="article:modified_time" content="2025-08-04T03:59:37+00:00"><link rel=canonical href=/posts/supabase-deep-dive/><link rel=preload href=/fonts/fa-brands-400.woff2 as=font type=font/woff2 crossorigin><link rel=preload href=/fonts/fa-regular-400.woff2 as=font type=font/woff2 crossorigin><link rel=preload href=/fonts/fa-solid-900.woff2 as=font type=font/woff2 crossorigin><link rel=stylesheet href=/css/coder.min.6445a802b9389c9660e1b07b724dcf5718b1065ed2d71b4eeaf981cc7cc5fc46.css integrity="sha256-ZEWoArk4nJZg4bB7ck3PVxixBl7S1xtO6vmBzHzF/EY=" crossorigin=anonymous media=screen><link rel=stylesheet href=/css/coder-dark.min.a00e6364bacbc8266ad1cc81230774a1397198f8cfb7bcba29b7d6fcb54ce57f.css integrity="sha256-oA5jZLrLyCZq0cyBIwd0oTlxmPjPt7y6KbfW/LVM5X8=" crossorigin=anonymous media=screen><link rel=icon type=image/svg+xml href=/images/favicon.svg sizes=any><link rel=icon type=image/png href=/images/favicon-32x32.png sizes=32x32><link rel=icon type=image/png href=/images/favicon-16x16.png sizes=16x16><link rel=apple-touch-icon href=/images/apple-touch-icon.png><link rel=apple-touch-icon sizes=180x180 href=/images/apple-touch-icon.png><link rel=manifest href=/site.webmanifest><link rel=mask-icon href=/images/safari-pinned-tab.svg color=#5bbad5></head><body class="preload-transitions colorscheme-auto"><div class=float-container><a id=dark-mode-toggle class=colorscheme-toggle><i class="fa-solid fa-adjust fa-fw" aria-hidden=true></i></a></div><main class=wrapper><nav class=navigation><section class=container><a class=navigation-title href=/>Eric X. Liu's Personal Page
+</a><input type=checkbox id=menu-toggle>
+<label class="menu-button float-right" for=menu-toggle><i class="fa-solid fa-bars fa-fw" aria-hidden=true></i></label><ul class=navigation-list><li class=navigation-item><a class=navigation-link href=/posts/>Posts</a></li><li class=navigation-item><a class=navigation-link href=https://chat.ericxliu.me>Chat</a></li><li class=navigation-item><a class=navigation-link href=https://git.ericxliu.me/user/oauth2/Authenitk>Git</a></li><li class=navigation-item><a class=navigation-link href=https://coder.ericxliu.me/api/v2/users/oidc/callback>Coder</a></li><li class=navigation-item><a class=navigation-link href=/>|</a></li><li class=navigation-item><a class=navigation-link href=https://sso.ericxliu.me>Sign in</a></li></ul></section></nav><div class=content><section class="container post"><article><header><div class=post-title><h1 class=title><a class=title-link href=/posts/supabase-deep-dive/>Supabase Deep Dive: It's Not Magic, It's Just Postgres</a></h1></div><div class=post-meta><div class=date><span class=posted-on><i class="fa-solid fa-calendar" aria-hidden=true></i>
+<time datetime=2025-08-03T00:00:00Z>August 3, 2025
+</time></span><span class=reading-time><i class="fa-solid fa-clock" aria-hidden=true></i>
+8-minute read</span></div></div></header><div class=post-content><p>In the world of Backend-as-a-Service (BaaS), platforms are often treated as magic boxes. You push data in, you get data out, and you hope the magic inside scales. While this simplicity is powerful, it can obscure the underlying mechanics, leaving developers wondering what&rsquo;s really going on.</p><p>Supabase enters this space with a radically different philosophy: <strong>transparency</strong>. It provides the convenience of a BaaS, but it’s built on the world&rsquo;s most trusted relational database: PostgreSQL. The &ldquo;magic&rdquo; isn&rsquo;t a proprietary black box; it&rsquo;s a carefully assembled suite of open-source tools that enhance Postgres, not hide it.</p><p>This deep dive will deconstruct that suite. We will move beyond the basics to explore the architectural patterns, security models, and development workflows that allow you to build robust, scalable applications. We will cover:</p><ul><li><strong>The Supabase Blueprint:</strong> A procedural guide to designing your application.</li><li><strong>The Pillars of Supabase:</strong> A detailed look at Auth, Storage, Functions, and Realtime.</li><li><strong>Transactional Realtime:</strong> How Supabase guarantees data consistency in a live environment.</li><li><strong>Best Practices:</strong> The practical knowledge you need before writing a single line of code.</li></ul><h3 id=the-guiding-philosophy-your-database-as-the-source-of-truth>The Guiding Philosophy: Your Database as the Source of Truth
+<a class=heading-link href=#the-guiding-philosophy-your-database-as-the-source-of-truth><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><p>The most critical shift when adopting Supabase is to see your database as more than just a data store. It is your <strong>single source of truth</strong>. This means your database schema is responsible for:</p><ul><li><strong>Structure:</strong> The tables and columns that define your data.</li><li><strong>Relationships:</strong> The foreign keys that link tables together.</li><li><strong>Integrity:</strong> The constraints (<code>NOT NULL</code>, <code>UNIQUE</code>) that ensure your data is always valid.</li><li><strong>Security:</strong> The access control rules that define who can do what.</li></ul><p>By leveraging PostgreSQL&rsquo;s native power, you get <strong>full ACID compliance</strong> (Atomicity, Consistency, Isolation, Durability) out of the box. You don&rsquo;t need to worry about application-level code to prevent orphan records or inconsistent states; the database guarantees it for you.</p><h3 id=the-supabase-design-blueprint-a-procedural-guide>The Supabase Design Blueprint: A Procedural Guide
+<a class=heading-link href=#the-supabase-design-blueprint-a-procedural-guide><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><p>To build a scalable application, follow a structured design process that moves from abstract ideas to concrete implementation.</p><h4 id=phase-1-conceptual-modeling-the-blueprint>Phase 1: Conceptual Modeling (The Blueprint)
+<a class=heading-link href=#phase-1-conceptual-modeling-the-blueprint><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h4><p>Before touching the Supabase dashboard, map out your application on paper.</p><ol><li><strong>Identify the &ldquo;Nouns&rdquo;:</strong> These are your core data objects, which will become your database tables. For a project management app, they are <code>projects</code>, <code>tasks</code>, <code>users</code>, <code>comments</code>.</li><li><strong>Define the &ldquo;Verbs&rdquo;:</strong> These are the user actions. &ldquo;A user <em>creates</em> a task.&rdquo; &ldquo;A user <em>assigns</em> a task to another user.&rdquo; These actions will inform your security policies and APIs.</li><li><strong>Map Relationships:</strong> How do the nouns connect? A <code>task</code> belongs to one <code>project</code>. A <code>user</code> can have many <code>tasks</code>. A <code>project</code> can have many <code>users</code> (a many-to-many relationship, requiring a <code>project_users</code> join table).</li></ol><h4 id=phase-2-the-foundation-schema--migrations>Phase 2: The Foundation (Schema & Migrations)
+<a class=heading-link href=#phase-2-the-foundation-schema--migrations><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h4><p>Translate your model into SQL. For any serious project, use the <strong>Supabase CLI</strong> to manage this process.</p><ol><li><strong>Develop Locally:</strong> Run a full Supabase stack on your machine with <code>supabase start</code>.</li><li><strong>Create Migration Files:</strong> Write your <code>CREATE TABLE</code> statements in SQL files. Define columns, data types, and foreign key <code>REFERENCES</code> to enforce your relationships.</li><li><strong>Version Control:</strong> Commit these migration files to Git. Your database schema is now version-controlled alongside your application code.</li><li><strong>Deploy:</strong> Use <code>supabase db push</code> to apply your migrations to your live production database. This workflow is safe, repeatable, and professional.</li></ol><h4 id=phase-3-the-security-layer-row-level-security>Phase 3: The Security Layer (Row Level Security)
+<a class=heading-link href=#phase-3-the-security-layer-row-level-security><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h4><p>This is not an optional step. RLS is the cornerstone of Supabase security.</p><ol><li><strong>Deny by Default:</strong> For any table holding user data, immediately enable RLS. This blocks all access until you explicitly grant it.</li></ol><div class=highlight><pre tabindex=0 style=color:#e6edf3;background-color:#0d1117;-moz-tab-size:4;-o-tab-size:4;tab-size:4><code class=language-sql data-lang=sql><span style=display:flex><span><span style=color:#ff7b72>ALTER</span><span style=color:#6e7681> </span><span style=color:#ff7b72>TABLE</span><span style=color:#6e7681> </span>tasks<span style=color:#6e7681> </span>ENABLE<span style=color:#6e7681> </span><span style=color:#ff7b72>ROW</span><span style=color:#6e7681> </span><span style=color:#ff7b72>LEVEL</span><span style=color:#6e7681> </span><span style=color:#ff7b72>SECURITY</span>;<span style=color:#6e7681>
+</span></span></span></code></pre></div><ol start=2><li><strong>Write &ldquo;Allow&rdquo; Policies:</strong> Create policies based on your user stories. Policies are SQL rules that the database enforces on every single query.</li></ol><div class=highlight><pre tabindex=0 style=color:#e6edf3;background-color:#0d1117;-moz-tab-size:4;-o-tab-size:4;tab-size:4><code class=language-sql data-lang=sql><span style=display:flex><span><span style=color:#8b949e;font-style:italic>-- Users can see tasks in projects they are a member of.
+</span></span></span><span style=display:flex><span><span style=color:#8b949e;font-style:italic></span><span style=color:#ff7b72>CREATE</span><span style=color:#6e7681> </span>POLICY<span style=color:#6e7681> </span><span style=color:#a5d6ff>&#34;Allow read access to tasks in user&#39;s projects&#34;</span><span style=color:#6e7681>
+</span></span></span><span style=display:flex><span><span style=color:#6e7681></span><span style=color:#ff7b72>ON</span><span style=color:#6e7681> </span>tasks<span style=color:#6e7681> </span><span style=color:#ff7b72>FOR</span><span style=color:#6e7681> </span><span style=color:#ff7b72>SELECT</span><span style=color:#6e7681>
+</span></span></span><span style=display:flex><span><span style=color:#6e7681></span><span style=color:#ff7b72>USING</span><span style=color:#6e7681> </span>(<span style=color:#6e7681>
+</span></span></span><span style=display:flex><span><span style=color:#6e7681>  </span><span style=color:#ff7b72>EXISTS</span><span style=color:#6e7681> </span>(<span style=color:#6e7681>
+</span></span></span><span style=display:flex><span><span style=color:#6e7681>    </span><span style=color:#ff7b72>SELECT</span><span style=color:#6e7681> </span><span style=color:#a5d6ff>1</span><span style=color:#6e7681> </span><span style=color:#ff7b72>FROM</span><span style=color:#6e7681> </span>project_users<span style=color:#6e7681>
+</span></span></span><span style=display:flex><span><span style=color:#6e7681>    </span><span style=color:#ff7b72>WHERE</span><span style=color:#6e7681> </span>project_users.project_id<span style=color:#6e7681> </span><span style=color:#ff7b72;font-weight:700>=</span><span style=color:#6e7681> </span>tasks.project_id<span style=color:#6e7681>
+</span></span></span><span style=display:flex><span><span style=color:#6e7681>    </span><span style=color:#ff7b72>AND</span><span style=color:#6e7681> </span>project_users.user_id<span style=color:#6e7681> </span><span style=color:#ff7b72;font-weight:700>=</span><span style=color:#6e7681> </span>auth.uid()<span style=color:#6e7681>
+</span></span></span><span style=display:flex><span><span style=color:#6e7681>  </span>)<span style=color:#6e7681>
+</span></span></span><span style=display:flex><span><span style=color:#6e7681></span>);<span style=color:#6e7681>
+</span></span></span><span style=display:flex><span><span style=color:#6e7681>
+</span></span></span><span style=display:flex><span><span style=color:#6e7681></span><span style=color:#8b949e;font-style:italic>-- Users can only insert tasks for themselves.
+</span></span></span><span style=display:flex><span><span style=color:#8b949e;font-style:italic></span><span style=color:#ff7b72>CREATE</span><span style=color:#6e7681> </span>POLICY<span style=color:#6e7681> </span><span style=color:#a5d6ff>&#34;Allow users to create their own tasks&#34;</span><span style=color:#6e7681>
+</span></span></span><span style=display:flex><span><span style=color:#6e7681></span><span style=color:#ff7b72>ON</span><span style=color:#6e7681> </span>tasks<span style=color:#6e7681> </span><span style=color:#ff7b72>FOR</span><span style=color:#6e7681> </span><span style=color:#ff7b72>INSERT</span><span style=color:#6e7681>
+</span></span></span><span style=display:flex><span><span style=color:#6e7681></span><span style=color:#ff7b72>WITH</span><span style=color:#6e7681> </span><span style=color:#ff7b72>CHECK</span><span style=color:#6e7681> </span>(<span style=color:#6e7681> </span>auth.uid()<span style=color:#6e7681> </span><span style=color:#ff7b72;font-weight:700>=</span><span style=color:#6e7681> </span>tasks.assignee_id<span style=color:#6e7681> </span>);<span style=color:#6e7681>
+</span></span></span></code></pre></div><p>The <code>auth.uid()</code> function is a special Supabase utility that securely returns the ID of the logged-in user making the request.</p><h4 id=phase-4-the-apis-data-access>Phase 4: The APIs (Data Access)
+<a class=heading-link href=#phase-4-the-apis-data-access><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h4><p>With your data structured and secured, you can now build the access points.</p><ul><li><strong>For Simple CRUD:</strong> Use Supabase&rsquo;s auto-generated API. It&rsquo;s convenient, respects all your RLS policies, and is perfect for simple reads and writes on a single table.</li></ul><div class=highlight><pre tabindex=0 style=color:#e6edf3;background-color:#0d1117;-moz-tab-size:4;-o-tab-size:4;tab-size:4><code class=language-javascript data-lang=javascript><span style=display:flex><span><span style=color:#ff7b72>const</span> { data, error } <span style=color:#ff7b72;font-weight:700>=</span> <span style=color:#ff7b72>await</span> supabase.from(<span style=color:#a5d6ff>&#39;tasks&#39;</span>).select(<span style=color:#a5d6ff>&#39;*&#39;</span>);
+</span></span></code></pre></div><ul><li><strong>For Complex Logic:</strong> Use PostgreSQL Functions (RPC). Encapsulate complex <code>JOIN</code>s or multi-step transactions into a single, callable function. This reduces network chattiness and keeps your business logic secure on the server.</li></ul><div class=highlight><pre tabindex=0 style=color:#e6edf3;background-color:#0d1117;-moz-tab-size:4;-o-tab-size:4;tab-size:4><code class=language-sql data-lang=sql><span style=display:flex><span><span style=color:#8b949e;font-style:italic>-- A function to get a task and its project name in one call
+</span></span></span><span style=display:flex><span><span style=color:#8b949e;font-style:italic></span><span style=color:#ff7b72>CREATE</span><span style=color:#6e7681> </span><span style=color:#ff7b72>OR</span><span style=color:#6e7681> </span><span style=color:#ff7b72>REPLACE</span><span style=color:#6e7681> </span><span style=color:#ff7b72>FUNCTION</span><span style=color:#6e7681> </span>get_task_with_project(task_id_input<span style=color:#6e7681> </span>int)<span style=color:#6e7681>
+</span></span></span><span style=display:flex><span><span style=color:#6e7681></span><span style=color:#ff7b72>RETURNS</span><span style=color:#6e7681> </span><span style=color:#ff7b72>TABLE</span><span style=color:#6e7681> </span>(task_title<span style=color:#6e7681> </span>text,<span style=color:#6e7681> </span>project_name<span style=color:#6e7681> </span>text)<span style=color:#6e7681> </span><span style=color:#ff7b72>AS</span><span style=color:#6e7681> </span><span style=color:#f85149>$$</span><span style=color:#6e7681>
+</span></span></span><span style=display:flex><span><span style=color:#6e7681></span><span style=color:#ff7b72>BEGIN</span><span style=color:#6e7681>
+</span></span></span><span style=display:flex><span><span style=color:#6e7681>  </span><span style=color:#ff7b72>RETURN</span><span style=color:#6e7681> </span>QUERY<span style=color:#6e7681>
+</span></span></span><span style=display:flex><span><span style=color:#6e7681>    </span><span style=color:#ff7b72>SELECT</span><span style=color:#6e7681> </span>tasks.title,<span style=color:#6e7681> </span>projects.name<span style=color:#6e7681>
+</span></span></span><span style=display:flex><span><span style=color:#6e7681>    </span><span style=color:#ff7b72>FROM</span><span style=color:#6e7681> </span>tasks<span style=color:#6e7681>
+</span></span></span><span style=display:flex><span><span style=color:#6e7681>    </span><span style=color:#ff7b72>JOIN</span><span style=color:#6e7681> </span>projects<span style=color:#6e7681> </span><span style=color:#ff7b72>ON</span><span style=color:#6e7681> </span>tasks.project_id<span style=color:#6e7681> </span><span style=color:#ff7b72;font-weight:700>=</span><span style=color:#6e7681> </span>projects.id<span style=color:#6e7681>
+</span></span></span><span style=display:flex><span><span style=color:#6e7681>    </span><span style=color:#ff7b72>WHERE</span><span style=color:#6e7681> </span>tasks.id<span style=color:#6e7681> </span><span style=color:#ff7b72;font-weight:700>=</span><span style=color:#6e7681> </span>task_id_input;<span style=color:#6e7681>
+</span></span></span><span style=display:flex><span><span style=color:#6e7681></span><span style=color:#ff7b72>END</span>;<span style=color:#6e7681>
+</span></span></span><span style=display:flex><span><span style=color:#6e7681></span><span style=color:#f85149>$$</span><span style=color:#6e7681> </span><span style=color:#ff7b72>LANGUAGE</span><span style=color:#6e7681> </span>plpgsql;<span style=color:#6e7681>
+</span></span></span></code></pre></div><div class=highlight><pre tabindex=0 style=color:#e6edf3;background-color:#0d1117;-moz-tab-size:4;-o-tab-size:4;tab-size:4><code class=language-javascript data-lang=javascript><span style=display:flex><span><span style=color:#8b949e;font-style:italic>// Called simply from the frontend
+</span></span></span><span style=display:flex><span><span style=color:#8b949e;font-style:italic></span><span style=color:#ff7b72>const</span> { data, error } <span style=color:#ff7b72;font-weight:700>=</span> <span style=color:#ff7b72>await</span> supabase.rpc(<span style=color:#a5d6ff>&#39;get_task_with_project&#39;</span>, { task_id_input<span style=color:#ff7b72;font-weight:700>:</span> <span style=color:#a5d6ff>123</span> });
+</span></span></code></pre></div><h3 id=a-tour-of-the-core-services>A Tour of the Core Services
+<a class=heading-link href=#a-tour-of-the-core-services><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><p>Beyond the database, Supabase provides a suite of essential tools.</p><h4 id=authentication>Authentication
+<a class=heading-link href=#authentication><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h4><p>A complete user management system that integrates directly with your database. When a user signs up, a corresponding entry is created in the managed <code>auth.users</code> table, which you can then reference in your own tables.</p><div class=highlight><pre tabindex=0 style=color:#e6edf3;background-color:#0d1117;-moz-tab-size:4;-o-tab-size:4;tab-size:4><code class=language-javascript data-lang=javascript><span style=display:flex><span><span style=color:#8b949e;font-style:italic>// Sign up a new user and handle social logins with ease
+</span></span></span><span style=display:flex><span><span style=color:#8b949e;font-style:italic></span><span style=color:#ff7b72>const</span> { data, error } <span style=color:#ff7b72;font-weight:700>=</span> <span style=color:#ff7b72>await</span> supabase.auth.signUp({ email, password });
+</span></span><span style=display:flex><span><span style=color:#ff7b72>const</span> { data, error } <span style=color:#ff7b72;font-weight:700>=</span> <span style=color:#ff7b72>await</span> supabase.auth.signInWithOAuth({ provider<span style=color:#ff7b72;font-weight:700>:</span> <span style=color:#a5d6ff>&#39;github&#39;</span> });
+</span></span></code></pre></div><h4 id=storage>Storage
+<a class=heading-link href=#storage><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h4><p>A simple, S3-compatible object store for managing files like user avatars or documents. It&rsquo;s integrated with Postgres and RLS, allowing you to write fine-grained access policies on files and folders (buckets).</p><div class=highlight><pre tabindex=0 style=color:#e6edf3;background-color:#0d1117;-moz-tab-size:4;-o-tab-size:4;tab-size:4><code class=language-javascript data-lang=javascript><span style=display:flex><span><span style=color:#8b949e;font-style:italic>// Upload a user avatar to a public &#39;avatars&#39; bucket
+</span></span></span><span style=display:flex><span><span style=color:#8b949e;font-style:italic></span><span style=color:#ff7b72>const</span> { error } <span style=color:#ff7b72;font-weight:700>=</span> <span style=color:#ff7b72>await</span> supabase.storage
+</span></span><span style=display:flex><span>  .from(<span style=color:#a5d6ff>&#39;avatars&#39;</span>)
+</span></span><span style=display:flex><span>  .upload(<span style=color:#a5d6ff>`public/</span><span style=color:#a5d6ff>${</span>userId<span style=color:#a5d6ff>}</span><span style=color:#a5d6ff>.png`</span>, file);
+</span></span></code></pre></div><h4 id=edge-functions-vs-database-functions>Edge Functions vs. Database Functions
+<a class=heading-link href=#edge-functions-vs-database-functions><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h4><p>It&rsquo;s critical to know when to use which.</p><ul><li><strong>Database Functions (SQL):</strong> For data-intensive logic <em>inside</em> your database.</li><li><strong>Edge Functions (TypeScript/Deno):</strong> For connecting to the outside world. Use them to call third-party APIs (like Stripe for payments) or run computations that are not well-suited for SQL. This is where you use your secret <code>service_role</code> key, as the function runs in a trusted server environment.</li></ul><h3 id=the-realtime-engine-a-pubsub-system-for-postgres>The Realtime Engine: A Pub/Sub System for Postgres
+<a class=heading-link href=#the-realtime-engine-a-pubsub-system-for-postgres><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><p>Supabase&rsquo;s Realtime engine is a powerful feature for building live, interactive experiences.</p><h4 id=how-it-works-logical-replication>How it Works: Logical Replication
+<a class=heading-link href=#how-it-works-logical-replication><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h4><p>It&rsquo;s not magic; it leverages a core PostgreSQL feature.</p><ol><li>When you enable Realtime on a table, Supabase creates a <strong>Publication</strong> for it.</li><li>The Realtime server subscribes to this publication via a <strong>Logical Replication Slot</strong>.</li><li>When a transaction is <strong>successfully committed</strong> to your database, the change is written to Postgres&rsquo;s Write-Ahead Log (WAL).</li><li>The WAL change is then sent to the Realtime server through the replication slot.</li><li>The server converts this database event into a JSON payload and broadcasts it over a WebSocket to all subscribed clients.</li></ol><h4 id=transactional-integrity>Transactional Integrity
+<a class=heading-link href=#transactional-integrity><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h4><p>The most important guarantee of this system is its relationship with database transactions. An event is <strong>only broadcast <em>after</em> a transaction is fully and successfully committed.</strong> If a transaction is rolled back due to an error, the replication slot receives nothing, and no Realtime event is ever sent. This means you can trust that every Realtime message you receive corresponds to data that is permanently and consistently stored in your database.</p><h4 id=use-cases-and-limitations>Use Cases and Limitations
+<a class=heading-link href=#use-cases-and-limitations><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h4><ul><li><strong>Use For:</strong> Small, JSON-based messages like chat messages, live notifications, activity feeds, and presence indicators (&ldquo;who&rsquo;s online&rdquo;). Use the <code>broadcast</code> feature for ephemeral data like cursor positions that you don&rsquo;t need to save.</li><li><strong>Do NOT Use For:</strong> Large, continuous data streams. It is <strong>not</strong> a replacement for WebRTC for video/audio calls. The system is designed for small, infrequent payloads.</li></ul><div class=highlight><pre tabindex=0 style=color:#e6edf3;background-color:#0d1117;-moz-tab-size:4;-o-tab-size:4;tab-size:4><code class=language-javascript data-lang=javascript><span style=display:flex><span><span style=color:#ff7b72>const</span> channel <span style=color:#ff7b72;font-weight:700>=</span> supabase.channel(<span style=color:#a5d6ff>&#39;public:messages&#39;</span>);
+</span></span><span style=display:flex><span>
+</span></span><span style=display:flex><span><span style=color:#8b949e;font-style:italic>// Subscribe to new rows in the &#39;messages&#39; table
+</span></span></span><span style=display:flex><span><span style=color:#8b949e;font-style:italic></span>channel
+</span></span><span style=display:flex><span>  .on(
+</span></span><span style=display:flex><span>    <span style=color:#a5d6ff>&#39;postgres_changes&#39;</span>,
+</span></span><span style=display:flex><span>    { event<span style=color:#ff7b72;font-weight:700>:</span> <span style=color:#a5d6ff>&#39;INSERT&#39;</span>, schema<span style=color:#ff7b72;font-weight:700>:</span> <span style=color:#a5d6ff>&#39;public&#39;</span>, table<span style=color:#ff7b72;font-weight:700>:</span> <span style=color:#a5d6ff>&#39;messages&#39;</span> },
+</span></span><span style=display:flex><span>    (payload) =&gt; {
+</span></span><span style=display:flex><span>      console.log(<span style=color:#a5d6ff>&#39;New message received!&#39;</span>, payload.<span style=color:#ff7b72>new</span>);
+</span></span><span style=display:flex><span>      <span style=color:#8b949e;font-style:italic>// Update your UI here
+</span></span></span><span style=display:flex><span><span style=color:#8b949e;font-style:italic></span>    }
+</span></span><span style=display:flex><span>  )
+</span></span><span style=display:flex><span>  .subscribe();
+</span></span></code></pre></div><h3 id=final-words-of-advice>Final Words of Advice
+<a class=heading-link href=#final-words-of-advice><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><ul><li><strong>Frontend Freedom:</strong> Supabase is frontend-agnostic, but meta-frameworks like <strong>Next.js</strong> and <strong>SvelteKit</strong> offer a &ldquo;golden path&rdquo; with Auth Helpers that simplify server-side rendering and data fetching.</li><li><strong>Embrace the CLI:</strong> Use the Supabase CLI for a professional, safe, and repeatable development workflow. Don&rsquo;t manage your production schema by clicking in the UI.</li><li><strong>Know Your Keys:</strong> Use the public <code>anon</code> key in the browser. Guard the secret <code>service_role</code> key and only use it in secure server environments like Edge Functions.</li><li><strong>Indexes Matter:</strong> For fast queries on large tables, <code>CREATE INDEX</code> on frequently queried columns. Performance is not automatic.</li></ul><p>By understanding these principles, you can leverage Supabase not as a simple BaaS, but as a powerful, transparent, and scalable platform for building next-generation applications on the solid foundation of PostgreSQL.</p></div><footer><div id=disqus_thread></div><script>window.disqus_config=function(){},function(){if(["localhost","127.0.0.1"].indexOf(window.location.hostname)!=-1){document.getElementById("disqus_thread").innerHTML="Disqus comments not available by default when the website is previewed locally.";return}var t=document,e=t.createElement("script");e.async=!0,e.src="//ericxliu-me.disqus.com/embed.js",e.setAttribute("data-timestamp",+new Date),(t.head||t.body).appendChild(e)}(),document.addEventListener("themeChanged",function(){document.readyState=="complete"&&DISQUS.reset({reload:!0,config:disqus_config})})</script></footer></article><link rel=stylesheet href=https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/katex.min.css integrity=sha384-vKruj+a13U8yHIkAyGgK1J3ArTLzrFGBbBc0tDp4ad/EyewESeXE/Iv67Aj8gKZ0 crossorigin=anonymous><script defer src=https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/katex.min.js integrity=sha384-PwRUT/YqbnEjkZO0zZxNqcxACrXe+j766U2amXcgMg5457rve2Y7I6ZJSm2A0mS4 crossorigin=anonymous></script><script defer src=https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/contrib/auto-render.min.js integrity=sha384-+VBxd3r6XgURycqtZ117nYw44OOcIax56Z4dCRWbxyPt0Koah1uHoK0o4+/RRE05 crossorigin=anonymous onload='renderMathInElement(document.body,{delimiters:[{left:"$$",right:"$$",display:!0},{left:"$",right:"$",display:!1},{left:"\\(",right:"\\)",display:!1},{left:"\\[",right:"\\]",display:!0}]})'></script></section></div><footer class=footer><section class=container>©
+2016 -
+2025
+Eric X. Liu
+<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/69c2890">[69c2890]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script><script defer src=https://static.cloudflareinsights.com/beacon.min.js data-cf-beacon='{"token": "987638e636ce4dbb932d038af74c17d1"}'></script></body></html>
--- a/posts/t5-the-transformer-that-zigged-when-others-zagged-an-architectural-deep-dive/index.html
+++ b/posts/t5-the-transformer-that-zigged-when-others-zagged-an-architectural-deep-dive/index.html
@@ -0,0 +1,33 @@
+<!doctype html><html lang=en><head><title>An Architectural Deep Dive of T5 · Eric X. Liu's Personal Page</title><meta charset=utf-8><meta name=viewport content="width=device-width,initial-scale=1"><meta name=color-scheme content="light dark"><meta name=author content="Eric X. Liu"><meta name=description content="In the rapidly evolving landscape of Large Language Models, a few key architectures define the dominant paradigms. Today, the &ldquo;decoder-only&rdquo; model, popularized by the GPT series and its successors like LLaMA and Mistral, reigns supreme. These models are scaled to incredible sizes and excel at in-context learning.
+But to truly understand the field, we must look at the pivotal models that explored different paths. Google&rsquo;s T5, or Text-to-Text Transfer Transformer, stands out as one of the most influential. It didn&rsquo;t just introduce a new model; it proposed a new philosophy. This article dives deep into the architecture of T5, how it fundamentally differs from modern LLMs, and the lasting legacy of its unique design choices."><meta name=keywords content="software engineer,performance engineering,Google engineer,tech blog,software development,performance optimization,Eric Liu,engineering blog,mountain biking,Jeep enthusiast,overlanding,camping,outdoor adventures"><meta name=twitter:card content="summary"><meta name=twitter:title content="An Architectural Deep Dive of T5"><meta name=twitter:description content="In the rapidly evolving landscape of Large Language Models, a few key architectures define the dominant paradigms. Today, the “decoder-only” model, popularized by the GPT series and its successors like LLaMA and Mistral, reigns supreme. These models are scaled to incredible sizes and excel at in-context learning.
+But to truly understand the field, we must look at the pivotal models that explored different paths. Google’s T5, or Text-to-Text Transfer Transformer, stands out as one of the most influential. It didn’t just introduce a new model; it proposed a new philosophy. This article dives deep into the architecture of T5, how it fundamentally differs from modern LLMs, and the lasting legacy of its unique design choices."><meta property="og:url" content="/posts/t5-the-transformer-that-zigged-when-others-zagged-an-architectural-deep-dive/"><meta property="og:site_name" content="Eric X. Liu's Personal Page"><meta property="og:title" content="An Architectural Deep Dive of T5"><meta property="og:description" content="In the rapidly evolving landscape of Large Language Models, a few key architectures define the dominant paradigms. Today, the “decoder-only” model, popularized by the GPT series and its successors like LLaMA and Mistral, reigns supreme. These models are scaled to incredible sizes and excel at in-context learning.
+But to truly understand the field, we must look at the pivotal models that explored different paths. Google’s T5, or Text-to-Text Transfer Transformer, stands out as one of the most influential. It didn’t just introduce a new model; it proposed a new philosophy. This article dives deep into the architecture of T5, how it fundamentally differs from modern LLMs, and the lasting legacy of its unique design choices."><meta property="og:locale" content="en"><meta property="og:type" content="article"><meta property="article:section" content="posts"><meta property="article:published_time" content="2025-06-01T00:00:00+00:00"><meta property="article:modified_time" content="2025-08-03T03:41:10+00:00"><link rel=canonical href=/posts/t5-the-transformer-that-zigged-when-others-zagged-an-architectural-deep-dive/><link rel=preload href=/fonts/fa-brands-400.woff2 as=font type=font/woff2 crossorigin><link rel=preload href=/fonts/fa-regular-400.woff2 as=font type=font/woff2 crossorigin><link rel=preload href=/fonts/fa-solid-900.woff2 as=font type=font/woff2 crossorigin><link rel=stylesheet href=/css/coder.min.6445a802b9389c9660e1b07b724dcf5718b1065ed2d71b4eeaf981cc7cc5fc46.css integrity="sha256-ZEWoArk4nJZg4bB7ck3PVxixBl7S1xtO6vmBzHzF/EY=" crossorigin=anonymous media=screen><link rel=stylesheet href=/css/coder-dark.min.a00e6364bacbc8266ad1cc81230774a1397198f8cfb7bcba29b7d6fcb54ce57f.css integrity="sha256-oA5jZLrLyCZq0cyBIwd0oTlxmPjPt7y6KbfW/LVM5X8=" crossorigin=anonymous media=screen><link rel=icon type=image/svg+xml href=/images/favicon.svg sizes=any><link rel=icon type=image/png href=/images/favicon-32x32.png sizes=32x32><link rel=icon type=image/png href=/images/favicon-16x16.png sizes=16x16><link rel=apple-touch-icon href=/images/apple-touch-icon.png><link rel=apple-touch-icon sizes=180x180 href=/images/apple-touch-icon.png><link rel=manifest href=/site.webmanifest><link rel=mask-icon href=/images/safari-pinned-tab.svg color=#5bbad5></head><body class="preload-transitions colorscheme-auto"><div class=float-container><a id=dark-mode-toggle class=colorscheme-toggle><i class="fa-solid fa-adjust fa-fw" aria-hidden=true></i></a></div><main class=wrapper><nav class=navigation><section class=container><a class=navigation-title href=/>Eric X. Liu's Personal Page
+</a><input type=checkbox id=menu-toggle>
+<label class="menu-button float-right" for=menu-toggle><i class="fa-solid fa-bars fa-fw" aria-hidden=true></i></label><ul class=navigation-list><li class=navigation-item><a class=navigation-link href=/posts/>Posts</a></li><li class=navigation-item><a class=navigation-link href=https://chat.ericxliu.me>Chat</a></li><li class=navigation-item><a class=navigation-link href=https://git.ericxliu.me/user/oauth2/Authenitk>Git</a></li><li class=navigation-item><a class=navigation-link href=https://coder.ericxliu.me/api/v2/users/oidc/callback>Coder</a></li><li class=navigation-item><a class=navigation-link href=/>|</a></li><li class=navigation-item><a class=navigation-link href=https://sso.ericxliu.me>Sign in</a></li></ul></section></nav><div class=content><section class="container post"><article><header><div class=post-title><h1 class=title><a class=title-link href=/posts/t5-the-transformer-that-zigged-when-others-zagged-an-architectural-deep-dive/>An Architectural Deep Dive of T5</a></h1></div><div class=post-meta><div class=date><span class=posted-on><i class="fa-solid fa-calendar" aria-hidden=true></i>
+<time datetime=2025-06-01T00:00:00Z>June 1, 2025
+</time></span><span class=reading-time><i class="fa-solid fa-clock" aria-hidden=true></i>
+6-minute read</span></div></div></header><div class=post-content><p>In the rapidly evolving landscape of Large Language Models, a few key architectures define the dominant paradigms. Today, the &ldquo;decoder-only&rdquo; model, popularized by the GPT series and its successors like LLaMA and Mistral, reigns supreme. These models are scaled to incredible sizes and excel at in-context learning.</p><p>But to truly understand the field, we must look at the pivotal models that explored different paths. Google&rsquo;s T5, or <strong>Text-to-Text Transfer Transformer</strong>, stands out as one of the most influential. It didn&rsquo;t just introduce a new model; it proposed a new philosophy. This article dives deep into the architecture of T5, how it fundamentally differs from modern LLMs, and the lasting legacy of its unique design choices.</p><h3 id=the-core-philosophy-everything-is-a-text-to-text-problem>The Core Philosophy: Everything is a Text-to-Text Problem
+<a class=heading-link href=#the-core-philosophy-everything-is-a-text-to-text-problem><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><p>The genius of T5 lies in its unifying framework. Instead of building different models or fine-tuning procedures for various NLP tasks, T5 reframes every task as a text-to-text problem. The model takes a string as input and generates a string as output, regardless of the underlying objective.</p><p>This is accomplished by adding a <strong>task prefix</strong> to the input. These prefixes are not conversational prompts like a GPT &ldquo;system prompt&rdquo;; they are learned triggers that the model is explicitly fine-tuned to recognize.</p><table><thead><tr><th style=text-align:left>Task</th><th style=text-align:left>T5 Input</th><th style=text-align:left>Expected T5 Output</th></tr></thead><tbody><tr><td style=text-align:left>Translation</td><td style=text-align:left><code>translate English to German: The cat is cute.</code></td><td style=text-align:left><code>Die Katze ist süß.</code></td></tr><tr><td style=text-align:left>Summarization</td><td style=text-align:left><code>summarize: [A long news article...]</code></td><td style=text-align:left><code>[A concise summary.]</code></td></tr><tr><td style=text-align:left>Classification</td><td style=text-align:left><code>cola sentence: The boys is walking.</code></td><td style=text-align:left><code>unacceptable</code></td></tr><tr><td style=text-align:left>Similarity</td><td style=text-align:left><code>stsb sentence1: The car is red. sentence2: The auto is crimson.</code></td><td style=text-align:left><code>4.8</code></td></tr></tbody></table><p>This elegant approach turns even classification into a generation task, where the model learns to generate the text of the correct label.</p><h3 id=the-engine-a-two-window-encoder-decoder-architecture>The Engine: A Two-Window Encoder-Decoder Architecture
+<a class=heading-link href=#the-engine-a-two-window-encoder-decoder-architecture><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><p>To execute this text-to-text mission, T5 uses the original Transformer&rsquo;s <strong>encoder-decoder architecture</strong>. This is the most significant point of divergence from modern decoder-only LLMs. The inference process works in two distinct stages:</p><h4 id=stage-1-the-encoder-the-understanding-window>Stage 1: The Encoder (The &ldquo;Understanding&rdquo; Window)
+<a class=heading-link href=#stage-1-the-encoder-the-understanding-window><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h4><p>When T5 receives an input like <code>summarize: [article text]</code>, the entire string is fed into the <strong>encoder</strong>.</p><ul><li><strong>Bidirectional Context:</strong> The encoder processes the input bidirectionally. Every token can see every other token in the input text simultaneously. This allows the model to build a deep, holistic understanding of the entire prompt and its context.</li><li><strong>Static Representation:</strong> The encoder&rsquo;s final output is not text. It&rsquo;s a set of numerical representations (hidden states) that encapsulates the meaning and intent of the input. This representation is generated once and remains static for the entire generation process.</li></ul><h4 id=stage-2-the-decoder-the-writing-window>Stage 2: The Decoder (The &ldquo;Writing&rdquo; Window)
+<a class=heading-link href=#stage-2-the-decoder-the-writing-window><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h4><p>The decoder is responsible for generating the output string token by token.</p><ul><li><strong>Autoregressive Generation:</strong> It begins with a <code>start-of-sequence</code> token and generates the output one word at a time.</li><li><strong>Cross-Attention:</strong> At each step, the decoder does two things: it looks at the text it has generated so far (its own &ldquo;decoder context&rdquo;), and crucially, it uses a mechanism called <strong>cross-attention</strong> to look back at the static representation created by the encoder. This allows the decoder&rsquo;s generation to be guided by the encoder&rsquo;s complete understanding of the prompt.</li><li><strong>Growing Context:</strong> The decoder&rsquo;s context window grows with each token it generates until it produces an <code>end-of-sequence</code> token, signaling that the task is complete.</li></ul><p>This two-window system is a powerful design, especially for tasks that require a full understanding of a source document before generating a new one (like translation or summarization).</p><h3 id=architectural-divergence-t5-vs-the-modern-llm-playbook>Architectural Divergence: T5 vs. The Modern LLM Playbook
+<a class=heading-link href=#architectural-divergence-t5-vs-the-modern-llm-playbook><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><p>Beyond its core architecture, T5 made several specific design choices that contrast with today&rsquo;s standards.</p><h4 id=1-positional-embeddings-relative-rpe-vs-rotary-rope>1. Positional Embeddings: Relative (RPE) vs. Rotary (RoPE)
+<a class=heading-link href=#1-positional-embeddings-relative-rpe-vs-rotary-rope><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h4><p>How a model knows the order of words is critical.</p><ul><li><strong>T5&rsquo;s Approach (RPE):</strong> T5 uses a form of <strong>Relative Positional Embedding</strong>. Instead of adding a position signal to the word embeddings, it adds a learned bias directly to the attention scores based on the relative distance between tokens. It&rsquo;s a clever way to encode position that is independent of sequence length.</li><li><strong>The Modern Standard (RoPE):</strong> Most modern LLMs (LLaMA, PaLM, Mistral) use <strong>Rotary Positional Embeddings</strong>. As detailed in the CS336 slides, RoPE works by mathematically <em>rotating</em> the Query and Key vectors based on their absolute position. This method has proven exceptionally effective for long sequences and is considered the current state-of-the-art.</li></ul><h4 id=2-the-feed-forward-network-an-extreme-experiment>2. The Feed-Forward Network: An Extreme Experiment
+<a class=heading-link href=#2-the-feed-forward-network-an-extreme-experiment><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h4><p>The Feed-Forward Network (FFN) inside each Transformer block is typically 4 times the model&rsquo;s hidden dimension (<code>d_model</code>). The original T5 11B model took a radical departure from this rule.</p><ul><li><strong>T5 11B&rsquo;s Choice:</strong> It used a small hidden dimension (<code>d_model = 1024</code>) but an astoundingly large FFN dimension (<code>d_ff = 65,536</code>), a <strong>64-times multiplier</strong>. The rationale was that modern accelerators (like Google&rsquo;s TPUs) are highly efficient at large, dense matrix multiplications.</li><li><strong>The Modern Standard:</strong> This experiment was not widely adopted. Later models, including T5&rsquo;s own successor <strong>T5 v1.1</strong>, reverted to the standard 4x multiplier (or ~2.66x when using GLU activations) for a better balance of parameters and performance.</li></ul><h4 id=3-denoising-span-corruption-vs-iterative-diffusion>3. Denoising: Span Corruption vs. Iterative Diffusion
+<a class=heading-link href=#3-denoising-span-corruption-vs-iterative-diffusion><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h4><p>While T5&rsquo;s pre-training is called &ldquo;denoising,&rdquo; it&rsquo;s conceptually different from the denoising in modern diffusion models.</p><ul><li><strong>T5&rsquo;s Denoising:</strong> This is <strong>span corruption</strong>. The model is shown a sentence with chunks of text masked out and learns to predict exactly what was removed in a single step. It&rsquo;s a fill-in-the-blanks task to learn rich language representations.</li><li><strong>Diffusion Denoising:</strong> This is a multi-step generative process. A clean text is gradually corrupted with noise, and the model learns to reverse this process step-by-step, allowing it to generate high-fidelity text from pure noise.</li></ul><h3 id=where-t5-was-ahead-of-its-time>Where T5 Was Ahead of its Time
+<a class=heading-link href=#where-t5-was-ahead-of-its-time><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><p>Despite its differences, the &ldquo;T5 v1.1&rdquo; variant pioneered several techniques that are now standard practice in the most advanced LLMs:</p><ul><li><strong>RMSNorm:</strong> It was one of the first major models to adopt Root Mean Square Normalization instead of LayerNorm, a choice now used by LLaMA, Mistral, and others for its efficiency and stability.</li><li><strong>Pre-Normalization:</strong> T5 applies the normalization layer <em>before</em> the attention and FFN blocks, a critical technique for enabling stable training of very deep networks.</li><li><strong>No Bias Terms:</strong> T5 v1.1 removed the bias parameters from its normalization and FFN layers, a small but important optimization for memory and stability that modern models follow.</li><li><strong>Gated Activations (GeGLU):</strong> While the original T5 used ReLU, T5 v1.1 adopted a Gated Linear Unit (GeGLU), presaging the move to GLU-family activations (like SwiGLU) that is now ubiquitous.</li></ul><h3 id=conclusion-the-lasting-legacy>Conclusion: The Lasting Legacy
+<a class=heading-link href=#conclusion-the-lasting-legacy><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><p>T5 represents a different evolutionary branch in the Transformer family tree. While the field has largely converged on the decoder-only architecture for its scalability in general-purpose models, T5&rsquo;s design remains a masterclass in purpose-built engineering.</p><p>Its text-to-text framework was revolutionary, its encoder-decoder structure is still a go-to for tasks like translation, and its refined T5 v1.1 architecture laid the groundwork for many of the stability and efficiency tricks we see in today&rsquo;s state-of-the-art models. T5 is more than just a model; it&rsquo;s a crucial case study in the architectural trade-offs that continue to shape the future of artificial intelligence.</p></div><footer><div id=disqus_thread></div><script>window.disqus_config=function(){},function(){if(["localhost","127.0.0.1"].indexOf(window.location.hostname)!=-1){document.getElementById("disqus_thread").innerHTML="Disqus comments not available by default when the website is previewed locally.";return}var t=document,e=t.createElement("script");e.async=!0,e.src="//ericxliu-me.disqus.com/embed.js",e.setAttribute("data-timestamp",+new Date),(t.head||t.body).appendChild(e)}(),document.addEventListener("themeChanged",function(){document.readyState=="complete"&&DISQUS.reset({reload:!0,config:disqus_config})})</script></footer></article><link rel=stylesheet href=https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/katex.min.css integrity=sha384-vKruj+a13U8yHIkAyGgK1J3ArTLzrFGBbBc0tDp4ad/EyewESeXE/Iv67Aj8gKZ0 crossorigin=anonymous><script defer src=https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/katex.min.js integrity=sha384-PwRUT/YqbnEjkZO0zZxNqcxACrXe+j766U2amXcgMg5457rve2Y7I6ZJSm2A0mS4 crossorigin=anonymous></script><script defer src=https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/contrib/auto-render.min.js integrity=sha384-+VBxd3r6XgURycqtZ117nYw44OOcIax56Z4dCRWbxyPt0Koah1uHoK0o4+/RRE05 crossorigin=anonymous onload='renderMathInElement(document.body,{delimiters:[{left:"$$",right:"$$",display:!0},{left:"$",right:"$",display:!1},{left:"\\(",right:"\\)",display:!1},{left:"\\[",right:"\\]",display:!0}]})'></script></section></div><footer class=footer><section class=container>©
+2016 -
+2025
+Eric X. Liu
+<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/69c2890">[69c2890]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script><script defer src=https://static.cloudflareinsights.com/beacon.min.js data-cf-beacon='{"token": "987638e636ce4dbb932d038af74c17d1"}'></script></body></html>
--- a/posts/transformer-s-core-mechanics/index.html
+++ b/posts/transformer-s-core-mechanics/index.html
@@ -0,0 +1,39 @@
+<!doctype html><html lang=en><head><title>Transformer's Core Mechanics · Eric X. Liu's Personal Page</title><meta charset=utf-8><meta name=viewport content="width=device-width,initial-scale=1"><meta name=color-scheme content="light dark"><meta name=author content="Eric X. Liu"><meta name=description content="The Transformer architecture is the bedrock of modern Large Language Models (LLMs). While its high-level success is widely known, a deeper understanding requires dissecting its core components. This article provides a detailed, technical breakdown of the fundamental concepts within a Transformer block, from the notion of &ldquo;channels&rdquo; to the intricate workings of the attention mechanism and its relationship with other advanced architectures like Mixture of Experts.
+
+  1. The &ldquo;Channel&rdquo;: A Foundational View of d_model
+  
+    
+    Link to heading
+  
+
+In deep learning, a &ldquo;channel&rdquo; can be thought of as a feature dimension. While this term is common in Convolutional Neural Networks for images (e.g., Red, Green, Blue channels), in LLMs, the analogous concept is the model&rsquo;s primary embedding dimension, commonly referred to as d_model."><meta name=keywords content="software engineer,performance engineering,Google engineer,tech blog,software development,performance optimization,Eric Liu,engineering blog,mountain biking,Jeep enthusiast,overlanding,camping,outdoor adventures"><meta name=twitter:card content="summary"><meta name=twitter:title content="Transformer's Core Mechanics"><meta name=twitter:description content="The Transformer architecture is the bedrock of modern Large Language Models (LLMs). While its high-level success is widely known, a deeper understanding requires dissecting its core components. This article provides a detailed, technical breakdown of the fundamental concepts within a Transformer block, from the notion of “channels” to the intricate workings of the attention mechanism and its relationship with other advanced architectures like Mixture of Experts.
+1. The “Channel”: A Foundational View of d_model Link to heading In deep learning, a “channel” can be thought of as a feature dimension. While this term is common in Convolutional Neural Networks for images (e.g., Red, Green, Blue channels), in LLMs, the analogous concept is the model’s primary embedding dimension, commonly referred to as d_model."><meta property="og:url" content="/posts/transformer-s-core-mechanics/"><meta property="og:site_name" content="Eric X. Liu's Personal Page"><meta property="og:title" content="Transformer's Core Mechanics"><meta property="og:description" content="The Transformer architecture is the bedrock of modern Large Language Models (LLMs). While its high-level success is widely known, a deeper understanding requires dissecting its core components. This article provides a detailed, technical breakdown of the fundamental concepts within a Transformer block, from the notion of “channels” to the intricate workings of the attention mechanism and its relationship with other advanced architectures like Mixture of Experts.
+1. The “Channel”: A Foundational View of d_model Link to heading In deep learning, a “channel” can be thought of as a feature dimension. While this term is common in Convolutional Neural Networks for images (e.g., Red, Green, Blue channels), in LLMs, the analogous concept is the model’s primary embedding dimension, commonly referred to as d_model."><meta property="og:locale" content="en"><meta property="og:type" content="article"><meta property="article:section" content="posts"><meta property="article:published_time" content="2025-04-01T00:00:00+00:00"><meta property="article:modified_time" content="2025-08-20T06:28:39+00:00"><link rel=canonical href=/posts/transformer-s-core-mechanics/><link rel=preload href=/fonts/fa-brands-400.woff2 as=font type=font/woff2 crossorigin><link rel=preload href=/fonts/fa-regular-400.woff2 as=font type=font/woff2 crossorigin><link rel=preload href=/fonts/fa-solid-900.woff2 as=font type=font/woff2 crossorigin><link rel=stylesheet href=/css/coder.min.6445a802b9389c9660e1b07b724dcf5718b1065ed2d71b4eeaf981cc7cc5fc46.css integrity="sha256-ZEWoArk4nJZg4bB7ck3PVxixBl7S1xtO6vmBzHzF/EY=" crossorigin=anonymous media=screen><link rel=stylesheet href=/css/coder-dark.min.a00e6364bacbc8266ad1cc81230774a1397198f8cfb7bcba29b7d6fcb54ce57f.css integrity="sha256-oA5jZLrLyCZq0cyBIwd0oTlxmPjPt7y6KbfW/LVM5X8=" crossorigin=anonymous media=screen><link rel=icon type=image/svg+xml href=/images/favicon.svg sizes=any><link rel=icon type=image/png href=/images/favicon-32x32.png sizes=32x32><link rel=icon type=image/png href=/images/favicon-16x16.png sizes=16x16><link rel=apple-touch-icon href=/images/apple-touch-icon.png><link rel=apple-touch-icon sizes=180x180 href=/images/apple-touch-icon.png><link rel=manifest href=/site.webmanifest><link rel=mask-icon href=/images/safari-pinned-tab.svg color=#5bbad5></head><body class="preload-transitions colorscheme-auto"><div class=float-container><a id=dark-mode-toggle class=colorscheme-toggle><i class="fa-solid fa-adjust fa-fw" aria-hidden=true></i></a></div><main class=wrapper><nav class=navigation><section class=container><a class=navigation-title href=/>Eric X. Liu's Personal Page
+</a><input type=checkbox id=menu-toggle>
+<label class="menu-button float-right" for=menu-toggle><i class="fa-solid fa-bars fa-fw" aria-hidden=true></i></label><ul class=navigation-list><li class=navigation-item><a class=navigation-link href=/posts/>Posts</a></li><li class=navigation-item><a class=navigation-link href=https://chat.ericxliu.me>Chat</a></li><li class=navigation-item><a class=navigation-link href=https://git.ericxliu.me/user/oauth2/Authenitk>Git</a></li><li class=navigation-item><a class=navigation-link href=https://coder.ericxliu.me/api/v2/users/oidc/callback>Coder</a></li><li class=navigation-item><a class=navigation-link href=/>|</a></li><li class=navigation-item><a class=navigation-link href=https://sso.ericxliu.me>Sign in</a></li></ul></section></nav><div class=content><section class="container post"><article><header><div class=post-title><h1 class=title><a class=title-link href=/posts/transformer-s-core-mechanics/>Transformer's Core Mechanics</a></h1></div><div class=post-meta><div class=date><span class=posted-on><i class="fa-solid fa-calendar" aria-hidden=true></i>
+<time datetime=2025-04-01T00:00:00Z>April 1, 2025
+</time></span><span class=reading-time><i class="fa-solid fa-clock" aria-hidden=true></i>
+7-minute read</span></div></div></header><div class=post-content><p>The Transformer architecture is the bedrock of modern Large Language Models (LLMs). While its high-level success is widely known, a deeper understanding requires dissecting its core components. This article provides a detailed, technical breakdown of the fundamental concepts within a Transformer block, from the notion of &ldquo;channels&rdquo; to the intricate workings of the attention mechanism and its relationship with other advanced architectures like Mixture of Experts.</p><h3 id=1-the-channel-a-foundational-view-of-d_model>1. The &ldquo;Channel&rdquo;: A Foundational View of <code>d_model</code>
+<a class=heading-link href=#1-the-channel-a-foundational-view-of-d_model><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><p>In deep learning, a &ldquo;channel&rdquo; can be thought of as a feature dimension. While this term is common in Convolutional Neural Networks for images (e.g., Red, Green, Blue channels), in LLMs, the analogous concept is the model&rsquo;s primary embedding dimension, commonly referred to as <code>d_model</code>.</p><p>An input text is first tokenized, and each token is mapped to a vector of size <code>d_model</code> (e.g., 4096). Each of the 4096 dimensions in this vector can be considered a &ldquo;channel,&rdquo; representing a different semantic or syntactic feature of the token.</p><p>As this data, represented by a tensor of shape <code>[batch_size, sequence_length, d_model]</code>, progresses through the layers of the Transformer, these channels are continuously transformed. However, a critical design choice is that the output dimension of every main sub-layer (like the attention block or the FFN block) is also <code>d_model</code>. This consistency is essential for enabling <strong>residual connections</strong>, where the input to a block is added to its output (<code>output = input + SubLayer(input)</code>). This technique is vital for training the extremely deep networks common today.</p><h3 id=2-the-building-blocks-dimensions-of-key-layers>2. The Building Blocks: Dimensions of Key Layers
+<a class=heading-link href=#2-the-building-blocks-dimensions-of-key-layers><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><p>A Transformer layer is primarily composed of two sub-layers: a Multi-Head Attention block and a position-wise Feed-Forward Network (FFN). The parameters for these are stored in several key weight matrices. Understanding their dimensions is crucial.</p><p>Let&rsquo;s define our variables:</p><ul><li><code>d_model</code>: The core embedding dimension.</li><li><code>d_ff</code>: The inner dimension of the FFN, typically <code>4 * d_model</code>.</li><li><code>h</code>: The number of attention heads.</li><li><code>d_head</code>: The dimension of each attention head, where <code>d_model = h * d_head</code>.</li></ul><p>The dimensions of the weight matrices are as follows:</p><table><thead><tr><th>Layer</th><th>Weight Matrix</th><th>Input Vector Shape</th><th>Output Vector Shape</th><th><strong>Weight Matrix Dimension</strong></th></tr></thead><tbody><tr><td><strong>Attention Projections</strong></td><td></td><td></td><td></td><td></td></tr><tr><td>Query</td><td><code>W_Q</code></td><td><code>d_model</code></td><td><code>d_model</code></td><td><strong><code>[d_model, d_model]</code></strong></td></tr><tr><td>Key</td><td><code>W_K</code></td><td><code>d_model</code></td><td><code>d_model</code></td><td><strong><code>[d_model, d_model]</code></strong></td></tr><tr><td>Value</td><td><code>W_V</code></td><td><code>d_model</code></td><td><code>d_model</code></td><td><strong><code>[d_model, d_model]</code></strong></td></tr><tr><td>Output</td><td><code>W_O</code></td><td><code>d_model</code></td><td><code>d_model</code></td><td><strong><code>[d_model, d_model]</code></strong></td></tr><tr><td><strong>Feed-Forward Network</strong></td><td></td><td></td><td></td><td></td></tr><tr><td>Layer 1 (Up-projection)</td><td><code>W_ff1</code></td><td><code>d_model</code></td><td><code>d_ff</code></td><td><strong><code>[d_model, d_ff]</code></strong></td></tr><tr><td>Layer 2 (Down-projection)</td><td><code>W_ff2</code></td><td><code>d_ff</code></td><td><code>d_model</code></td><td><strong><code>[d_ff, d_model]</code></strong></td></tr></tbody></table><h3 id=3-deconstructing-multi-head-attention-mha>3. Deconstructing Multi-Head Attention (MHA)
+<a class=heading-link href=#3-deconstructing-multi-head-attention-mha><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><p>The core innovation of the Transformer is Multi-Head Attention. It allows the model to weigh the importance of different tokens in the sequence from multiple perspectives simultaneously.
+<img src=/images/transformer-s-core-mechanics/.png alt></p><h4 id=31-the-why-beyond-a-single-attention>3.1. The &ldquo;Why&rdquo;: Beyond a Single Attention
+<a class=heading-link href=#31-the-why-beyond-a-single-attention><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h4><p>A single attention mechanism would force the model to average all types of linguistic relationships into one pattern. MHA avoids this by creating <code>h</code> parallel subspaces. Each &ldquo;head&rdquo; can specialize, with one head learning syntactic dependencies, another tracking semantic similarity, and so on. This creates a much richer representation.</p><h4 id=32-an-encodingdecoding-analogy>3.2. An Encoding/Decoding Analogy
+<a class=heading-link href=#32-an-encodingdecoding-analogy><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h4><p>A powerful way to conceptualize the attention calculation is as a two-stage process:</p><ol><li><strong>Encoding Relationships:</strong> The first part of the calculation, <code>softmax(Q @ K.T)</code>, can be seen as an encoding step. It does not use the actual &ldquo;content&rdquo; of the tokens (the <code>V</code> vectors). Instead, it uses the Queries and Keys to build a dynamic &ldquo;relationship map&rdquo; between tokens in the sequence. This map, a matrix of attention scores, answers the question: &ldquo;For each token, how important is every other token right now?&rdquo;</li><li><strong>Decoding via Information Retrieval:</strong> The second part, <code>scores @ V</code>, acts as a decoding step. It uses the relationship map to retrieve and synthesize information. For each token, it creates a new vector by taking a weighted sum of all the <code>V</code> vectors in the sequence, using the scores as the precise mixing recipe. It decodes the relational structure into a new, context-aware representation.</li></ol><h4 id=33-the-how-a-step-by-step-flow>3.3. The &ldquo;How&rdquo;: A Step-by-Step Flow
+<a class=heading-link href=#33-the-how-a-step-by-step-flow><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h4><p>The MHA process is designed for maximum computational efficiency.</p><ol><li><strong>Initial Projections:</strong> The input vectors (shape <code>[seq_len, d_model]</code>) are multiplied by <code>W_Q</code>, <code>W_K</code>, and <code>W_V</code>. These matrices are all <code>[d_model, d_model]</code> not to create one large query, but to <strong>efficiently compute the vectors for all <code>h</code> heads at once</strong>. The single large output vector is then reshaped into <code>h</code> separate vectors, each of size <code>d_head</code>.</li><li><strong>Attention Score Calculation:</strong> For each head <code>i</code>, a score matrix is calculated: <code>scores_i = softmax( (Q_i @ K_i.T) / sqrt(d_head) )</code>. Note that <code>Q_i</code> and <code>K_i</code> have dimensions <code>[seq_len, d_head]</code>, so the resulting <code>scores_i</code> matrix has a dimension of <strong><code>[seq_len, seq_len]</code></strong>.</li><li><strong>Weighted Value Calculation:</strong> The scores are used to create a weighted sum of the Value vectors for each head: <code>output_i = scores_i @ V_i</code>. Since <code>scores_i</code> is <code>[seq_len, seq_len]</code> and <code>V_i</code> is <code>[seq_len, d_head]</code>, the resulting <code>output_i</code> has a dimension of <strong><code>[seq_len, d_head]</code></strong>. This is the final output of a single head.</li><li><strong>Concatenation and Final Projection:</strong> The outputs of all <code>h</code> heads are concatenated along the last dimension. This produces a single large matrix of shape <code>[seq_len, h * d_head]</code>, which is equivalent to <code>[seq_len, d_model]</code>. This matrix is then passed through the final output projection layer, <code>W_O</code> (shape <code>[d_model, d_model]</code>), to produce the attention block&rsquo;s final output. The <code>W_O</code> matrix learns the optimal way to mix the information from all the specialized heads into a single, unified representation.</li></ol><h3 id=4-optimizing-attention-gqa-and-mqa>4. Optimizing Attention: GQA and MQA
+<a class=heading-link href=#4-optimizing-attention-gqa-and-mqa><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><p>During inference, storing the Key and Value vectors for all previous tokens (the KV Cache) is a major memory bottleneck. <strong>Grouped-Query Attention (GQA)</strong> and <strong>Multi-Query Attention (MQA)</strong> are architectural modifications that address this by allowing multiple Query heads to share the same Key and Value heads.</p><p>Let&rsquo;s use a concrete example, similar to Llama 2 7B:</p><ul><li><code>d_model</code> = 4096</li><li><code>h</code> = 32 Q heads</li><li><code>d_head</code> = 128</li><li><code>g</code> = 8 KV head groups for GQA</li></ul><p>The key insight is that only the dimensions of the <code>W_K</code> and <code>W_V</code> matrices change, which in turn reduces the size of the KV cache. The <code>W_Q</code> and <code>W_O</code> matrices remain <code>[4096, 4096]</code>.</p><table><thead><tr><th>Attention Type</th><th>No. of Q Heads</th><th>No. of KV Heads</th><th><code>W_K</code> & <code>W_V</code> Dimension</th><th>Relative KV Cache Size</th></tr></thead><tbody><tr><td><strong>MHA</strong> (Multi-Head)</td><td>32</td><td>32</td><td><code>[4096, 32*128]</code> = <code>[4096, 4096]</code></td><td>1x (Baseline)</td></tr><tr><td><strong>GQA</strong> (Grouped)</td><td>32</td><td>8</td><td><code>[4096, 8*128]</code> = <code>[4096, 1024]</code></td><td>1/4x</td></tr><tr><td><strong>MQA</strong> (Multi-Query)</td><td>32</td><td>1</td><td><code>[4096, 1*128]</code> = <code>[4096, 128]</code></td><td>1/32x</td></tr></tbody></table><p>GQA provides a robust balance, significantly reducing the memory and bandwidth requirements for the KV cache with negligible impact on model performance, making it a popular choice in modern LLMs.</p><h3 id=5-mha-vs-mixture-of-experts-moe-a-clarification>5. MHA vs. Mixture of Experts (MoE): A Clarification
+<a class=heading-link href=#5-mha-vs-mixture-of-experts-moe-a-clarification><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><p>While both MHA and MoE use the concept of &ldquo;experts,&rdquo; they are functionally and architecturally distinct.</p><ul><li><strong>MHA:</strong> The &ldquo;experts&rdquo; are the <strong>attention heads</strong>. All heads are active for every token to build a rich representation within the attention layer. This is akin to a board meeting where every member analyzes and contributes to every decision.</li><li><strong>MoE:</strong> The &ldquo;experts&rdquo; are full <strong>Feed-Forward Networks</strong>. A routing network selects a small subset of these FFNs for each token. This is a scaling strategy to increase a model&rsquo;s parameter count for greater capacity while keeping the computational cost fixed. It replaces the standard FFN block, whereas MHA <em>is</em> the attention block.</li></ul><p>By understanding these technical details, from the basic concept of a channel to the sophisticated interplay of heads and experts, one can build a more complete and accurate mental model of how LLMs truly operate.</p><hr><h3 id=references>References
+<a class=heading-link href=#references><i class="fa-solid fa-link" aria-hidden=true title="Link to heading"></i>
+<span class=sr-only>Link to heading</span></a></h3><ol><li>Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., &mldr; & Polosukhin, I. (2017). Attention is all you need. <em>Advances in neural information processing systems</em>, 30.</li><li>Shazeer, N., Mirhoseini, A., Maziarz, K., Davis, A., Le, Q., Hinton, G., & Dean, J. (2017). Outrageously large neural networks: The sparsely-gated mixture-of-experts layer. <em>arXiv preprint arXiv:1701.06538</em>.</li><li>Ainslie, J., Ontanon, J., Cakka, E., Dosovitskiy, A., & Le, Q. V. (2023). GQA: Training Generalized Multi-Query Transformer Models from Multi-Head Checkpoints. <em>arXiv preprint arXiv:2305.13245</em>.</li></ol></div><footer><div id=disqus_thread></div><script>window.disqus_config=function(){},function(){if(["localhost","127.0.0.1"].indexOf(window.location.hostname)!=-1){document.getElementById("disqus_thread").innerHTML="Disqus comments not available by default when the website is previewed locally.";return}var t=document,e=t.createElement("script");e.async=!0,e.src="//ericxliu-me.disqus.com/embed.js",e.setAttribute("data-timestamp",+new Date),(t.head||t.body).appendChild(e)}(),document.addEventListener("themeChanged",function(){document.readyState=="complete"&&DISQUS.reset({reload:!0,config:disqus_config})})</script></footer></article><link rel=stylesheet href=https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/katex.min.css integrity=sha384-vKruj+a13U8yHIkAyGgK1J3ArTLzrFGBbBc0tDp4ad/EyewESeXE/Iv67Aj8gKZ0 crossorigin=anonymous><script defer src=https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/katex.min.js integrity=sha384-PwRUT/YqbnEjkZO0zZxNqcxACrXe+j766U2amXcgMg5457rve2Y7I6ZJSm2A0mS4 crossorigin=anonymous></script><script defer src=https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/contrib/auto-render.min.js integrity=sha384-+VBxd3r6XgURycqtZ117nYw44OOcIax56Z4dCRWbxyPt0Koah1uHoK0o4+/RRE05 crossorigin=anonymous onload='renderMathInElement(document.body,{delimiters:[{left:"$$",right:"$$",display:!0},{left:"$",right:"$",display:!1},{left:"\\(",right:"\\)",display:!1},{left:"\\[",right:"\\]",display:!0}]})'></script></section></div><footer class=footer><section class=container>©
+2016 -
+2025
+Eric X. Liu
+<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/69c2890">[69c2890]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script><script defer src=https://static.cloudflareinsights.com/beacon.min.js data-cf-beacon='{"token": "987638e636ce4dbb932d038af74c17d1"}'></script></body></html>
--- a/posts/useful/index.html
+++ b/posts/useful/index.html
@@ -0,0 +1,12 @@
+<!doctype html><html lang=en><head><title>Some useful files · Eric X. Liu's Personal Page</title><meta charset=utf-8><meta name=viewport content="width=device-width,initial-scale=1"><meta name=color-scheme content="light dark"><meta name=author content="Eric X. Liu"><meta name=description content="
+rootCA.pem
+"><meta name=keywords content="software engineer,performance engineering,Google engineer,tech blog,software development,performance optimization,Eric Liu,engineering blog,mountain biking,Jeep enthusiast,overlanding,camping,outdoor adventures"><meta name=twitter:card content="summary"><meta name=twitter:title content="Some useful files"><meta name=twitter:description content="rootCA.pem"><meta property="og:url" content="/posts/useful/"><meta property="og:site_name" content="Eric X. Liu's Personal Page"><meta property="og:title" content="Some useful files"><meta property="og:description" content="rootCA.pem"><meta property="og:locale" content="en"><meta property="og:type" content="article"><meta property="article:section" content="posts"><meta property="article:published_time" content="2020-10-26T04:14:43+00:00"><meta property="article:modified_time" content="2025-08-03T08:37:28-07:00"><link rel=canonical href=/posts/useful/><link rel=preload href=/fonts/fa-brands-400.woff2 as=font type=font/woff2 crossorigin><link rel=preload href=/fonts/fa-regular-400.woff2 as=font type=font/woff2 crossorigin><link rel=preload href=/fonts/fa-solid-900.woff2 as=font type=font/woff2 crossorigin><link rel=stylesheet href=/css/coder.min.6445a802b9389c9660e1b07b724dcf5718b1065ed2d71b4eeaf981cc7cc5fc46.css integrity="sha256-ZEWoArk4nJZg4bB7ck3PVxixBl7S1xtO6vmBzHzF/EY=" crossorigin=anonymous media=screen><link rel=stylesheet href=/css/coder-dark.min.a00e6364bacbc8266ad1cc81230774a1397198f8cfb7bcba29b7d6fcb54ce57f.css integrity="sha256-oA5jZLrLyCZq0cyBIwd0oTlxmPjPt7y6KbfW/LVM5X8=" crossorigin=anonymous media=screen><link rel=icon type=image/svg+xml href=/images/favicon.svg sizes=any><link rel=icon type=image/png href=/images/favicon-32x32.png sizes=32x32><link rel=icon type=image/png href=/images/favicon-16x16.png sizes=16x16><link rel=apple-touch-icon href=/images/apple-touch-icon.png><link rel=apple-touch-icon sizes=180x180 href=/images/apple-touch-icon.png><link rel=manifest href=/site.webmanifest><link rel=mask-icon href=/images/safari-pinned-tab.svg color=#5bbad5></head><body class="preload-transitions colorscheme-auto"><div class=float-container><a id=dark-mode-toggle class=colorscheme-toggle><i class="fa-solid fa-adjust fa-fw" aria-hidden=true></i></a></div><main class=wrapper><nav class=navigation><section class=container><a class=navigation-title href=/>Eric X. Liu's Personal Page
+</a><input type=checkbox id=menu-toggle>
+<label class="menu-button float-right" for=menu-toggle><i class="fa-solid fa-bars fa-fw" aria-hidden=true></i></label><ul class=navigation-list><li class=navigation-item><a class=navigation-link href=/posts/>Posts</a></li><li class=navigation-item><a class=navigation-link href=https://chat.ericxliu.me>Chat</a></li><li class=navigation-item><a class=navigation-link href=https://git.ericxliu.me/user/oauth2/Authenitk>Git</a></li><li class=navigation-item><a class=navigation-link href=https://coder.ericxliu.me/api/v2/users/oidc/callback>Coder</a></li><li class=navigation-item><a class=navigation-link href=/>|</a></li><li class=navigation-item><a class=navigation-link href=https://sso.ericxliu.me>Sign in</a></li></ul></section></nav><div class=content><section class="container post"><article><header><div class=post-title><h1 class=title><a class=title-link href=/posts/useful/>Some useful files</a></h1></div><div class=post-meta><div class=date><span class=posted-on><i class="fa-solid fa-calendar" aria-hidden=true></i>
+<time datetime=2020-10-26T04:14:43Z>October 26, 2020
+</time></span><span class=reading-time><i class="fa-solid fa-clock" aria-hidden=true></i>
+One-minute read</span></div></div></header><div class=post-content><ul><li><a href=/rootCA.crt>rootCA.pem</a></li></ul></div><footer><div id=disqus_thread></div><script>window.disqus_config=function(){},function(){if(["localhost","127.0.0.1"].indexOf(window.location.hostname)!=-1){document.getElementById("disqus_thread").innerHTML="Disqus comments not available by default when the website is previewed locally.";return}var t=document,e=t.createElement("script");e.async=!0,e.src="//ericxliu-me.disqus.com/embed.js",e.setAttribute("data-timestamp",+new Date),(t.head||t.body).appendChild(e)}(),document.addEventListener("themeChanged",function(){document.readyState=="complete"&&DISQUS.reset({reload:!0,config:disqus_config})})</script></footer></article><link rel=stylesheet href=https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/katex.min.css integrity=sha384-vKruj+a13U8yHIkAyGgK1J3ArTLzrFGBbBc0tDp4ad/EyewESeXE/Iv67Aj8gKZ0 crossorigin=anonymous><script defer src=https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/katex.min.js integrity=sha384-PwRUT/YqbnEjkZO0zZxNqcxACrXe+j766U2amXcgMg5457rve2Y7I6ZJSm2A0mS4 crossorigin=anonymous></script><script defer src=https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/contrib/auto-render.min.js integrity=sha384-+VBxd3r6XgURycqtZ117nYw44OOcIax56Z4dCRWbxyPt0Koah1uHoK0o4+/RRE05 crossorigin=anonymous onload='renderMathInElement(document.body,{delimiters:[{left:"$$",right:"$$",display:!0},{left:"$",right:"$",display:!1},{left:"\\(",right:"\\)",display:!1},{left:"\\[",right:"\\]",display:!0}]})'></script></section></div><footer class=footer><section class=container>©
+2016 -
+2025
+Eric X. Liu
+<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/69c2890">[69c2890]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script><script defer src=https://static.cloudflareinsights.com/beacon.min.js data-cf-beacon='{"token": "987638e636ce4dbb932d038af74c17d1"}'></script></body></html>
--- a/robots.txt
+++ b/robots.txt
@@ -0,0 +1 @@
+User-agent: *
--- a/rootCA.crt
+++ b/rootCA.crt
@@ -0,0 +1,34 @@
+-----BEGIN CERTIFICATE-----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+-----END CERTIFICATE-----
--- a/site.webmanifest
+++ b/site.webmanifest
@@ -0,0 +1,20 @@
+{
+  "name": "Eric X. Liu's Personal Page",
+  "short_name": "Eric Liu",
+  "description": "Software & Performance Engineer at Google. Sharing insights about software engineering, performance optimization, tech industry experiences, mountain biking adventures, Jeep overlanding, and outdoor activities.",
+  "start_url": "/",
+  "display": "standalone",
+  "background_color": "#ffffff",
+  "theme_color": "#000000",
+  "icons": [
+    {
+      "src": "/images/gravatar.png",
+      "sizes": "192x192",
+      "type": "image/png",
+      "purpose": "any maskable"
+    }
+  ],
+  "categories": ["technology", "engineering", "blog"],
+  "lang": "en",
+  "orientation": "portrait-primary"
+}
--- a/sitemap.xml
+++ b/sitemap.xml
@@ -0,0 +1 @@
+<?xml version="1.0" encoding="utf-8" standalone="yes"?><urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9" xmlns:xhtml="http://www.w3.org/1999/xhtml"><url><loc>/</loc><lastmod>2025-08-20T06:28:39+00:00</lastmod><changefreq>weekly</changefreq><priority>0.5</priority></url><url><loc>/posts/</loc><lastmod>2025-08-20T06:28:39+00:00</lastmod><changefreq>weekly</changefreq><priority>0.5</priority></url><url><loc>/posts/quantization-in-llms/</loc><lastmod>2025-08-20T06:02:35+00:00</lastmod><changefreq>weekly</changefreq><priority>0.5</priority></url><url><loc>/posts/breville-barista-pro-maintenance/</loc><lastmod>2025-08-20T06:04:36+00:00</lastmod><changefreq>weekly</changefreq><priority>0.5</priority></url><url><loc>/posts/secure-boot-dkms-and-mok-on-proxmox-debian/</loc><lastmod>2025-08-14T06:50:22+00:00</lastmod><changefreq>weekly</changefreq><priority>0.5</priority></url><url><loc>/posts/how-rvq-teaches-llms-to-see-and-hear/</loc><lastmod>2025-08-08T17:36:52+00:00</lastmod><changefreq>weekly</changefreq><priority>0.5</priority></url><url><loc>/posts/supabase-deep-dive/</loc><lastmod>2025-08-04T03:59:37+00:00</lastmod><changefreq>weekly</changefreq><priority>0.5</priority></url><url><loc>/posts/ppo-for-language-models/</loc><lastmod>2025-08-20T06:04:36+00:00</lastmod><changefreq>weekly</changefreq><priority>0.5</priority></url><url><loc>/posts/mixture-of-experts-moe-models-challenges-solutions-in-practice/</loc><lastmod>2025-08-03T06:02:48+00:00</lastmod><changefreq>weekly</changefreq><priority>0.5</priority></url><url><loc>/posts/t5-the-transformer-that-zigged-when-others-zagged-an-architectural-deep-dive/</loc><lastmod>2025-08-03T03:41:10+00:00</lastmod><changefreq>weekly</changefreq><priority>0.5</priority></url><url><loc>/posts/espresso-theory-application-a-guide-for-the-breville-barista-pro/</loc><lastmod>2025-08-03T04:20:20+00:00</lastmod><changefreq>weekly</changefreq><priority>0.5</priority></url><url><loc>/posts/transformer-s-core-mechanics/</loc><lastmod>2025-08-20T06:28:39+00:00</lastmod><changefreq>weekly</changefreq><priority>0.5</priority></url><url><loc>/posts/useful/</loc><lastmod>2025-08-03T08:37:28-07:00</lastmod><changefreq>weekly</changefreq><priority>0.5</priority></url><url><loc>/about/</loc><lastmod>2020-06-16T23:30:17-07:00</lastmod><changefreq>weekly</changefreq><priority>0.5</priority></url><url><loc>/categories/</loc><changefreq>weekly</changefreq><priority>0.5</priority></url><url><loc>/tags/</loc><changefreq>weekly</changefreq><priority>0.5</priority></url></urlset>
--- a/static/images/gravatar.png
+++ b/static/images/gravatar.png
--- a/static/rootCA.pem
+++ b/static/rootCA.pem
@@ -1,33 +0,0 @@
-----BEGIN CERTIFICATE-----
-MIIFoDCCA4igAwIBAgIUJzlDGIEJdOQ0Shd1P0RJP5aangAwDQYJKoZIhvcNAQEL
-BQAwYTELMAkGA1UEBhMCVVMxCzAJBgNVBAgMAkNBMREwDwYDVQQHDAhTYW4gSm9z
-ZTEUMBIGA1UECgwLZXJpY3hsaXUubWUxHDAaBgkqhkiG9w0BCQEWDV9AZXJpY3hs
-aXUubWUwHhcNMjExMTIyMTgzMzQyWhcNMjQwOTExMTgzMzQyWjBhMQswCQYDVQQG
-EwJVUzELMAkGA1UECAwCQ0ExETAPBgNVBAcMCFNhbiBKb3NlMRQwEgYDVQQKDAtl
-cmljeGxpdS5tZTEcMBoGCSqGSIb3DQEJARYNX0BlcmljeGxpdS5tZTCCAiIwDQYJ
-KoZIhvcNAQEBBQADggIPADCCAgoCggIBAN50NMF7T6pFXWv6RG8xnMWR6eyIoykv
-hP1tFsBs0V9LxC58en3AmgyHnFbhc8m2qRh1ieeh59plL1r0MwePDX5/uUvS/Mh/
-NI7klZNlBICyLTJ6oOePWXxKFis9m9BH41SG32oDO1uEhwHLcUVU4UUxmNlpgjAX
-ULdAG/mHb/J+DDS5p3+s4TDUbVn/GWPsr6u+bn36/N24BJ8Vjo0wbQ8XNjWNGdZn
-h2KKpYkQzTYQB6+ixtZeDUIVGSh76zTzxBeXhZkvGke+yIp28qNpopetCJ4BgvYq
-XcM7OI3s/rn2rQGTa2bKnQHzVBlXbiaGDlYK7Cu16I4nPnW9yT9P6FYMUqM/Br4a
-MamoaJeMkwF9L8zIYx+Z2c91wx4pCjYzHiGmX+HXXrEVjrRPO2j88jgpvhUFMqQ3
-SG7sEsNJTiXw+SJJIB3QTYrODg/0FCC9Vz8GGhHugp5t/ZMJ5NU36AL0dGwF3MED
-ea2nrRB31cEUCRsOpmqe6wDwoQbXRrPhxdowiiECxGH0BXPi0fFM6rZ2DD0ATwQR
-9DOBwxCXiFXhORgW5sK3+CbXrIS+w8EwWOm0u7hLuY6lJ1onKU5dVwxno+B4YCWt
-u5i6duVi13IpXI1UV5dxrpE8BsO2FZKqtvOKAK9J6qFGk+YSrmnnUEUYD5ArZvUG
-Pt0XTKIM7nNvAgMBAAGjUDBOMB0GA1UdDgQWBBQEK7HddEflCZ9DL9VEIBXzB9dQ
-FTAfBgNVHSMEGDAWgBQEK7HddEflCZ9DL9VEIBXzB9dQFTAMBgNVHRMEBTADAQH/
-MA0GCSqGSIb3DQEBCwUAA4ICAQDKVGKjd1v6vecfNLZZ4+bqw4nwzzVwOdOWb2j+
-zqPmYT/ZzCCxeiWLIaYtOQWXR4eSzULWYAGauecVlVYydbRbDC6LXp/1NrfQuNpp
-6kd9JRGGdnNrW+0tEfJiXnEpOTwKncI1u6B0pvND8Gy6sxgjamyiKAh1vy0IZYJk
-2T7PXxljqGxFZXZ5Ese/ogPn5KRGPkOmbW/BQXWC//3Qe39J6lxy2/HwfZ9pa+AQ
-TxcJ/2OiDgBprMPJrHBiqvjoI9kp8vk3JhAQmbVM+8bpAIiiW8dPiEBDtROe/Wk5
-UuiebFQNIebaIG+nEruUR28Df3Q52k6dY4MWLVNqB9lKKCqnbYtxDUIQrFCSHAEq
-IdeOTEPjpkBr6UWwEunk32Mq6mdqmf5zBNaS64Wva43SLx+p/MIIacCYxOH7CHJX
-r6XO/tR95cO4N3LdA/aJYpY0M35tFftFKI/AD5vEwshgYDw9QU1fu3Wljw3wYSVx
-8YPKKwRkEBslEBmqf9YooDtGw3bLkQbJml0uMgxXOYI/VD95azvguq1lmcSdTTPu
-f1GC0YnpQnXT6gPHNLoMhGiQUTlwHp2GKdaW0Xb9DEOLurzBZ9FIQsvrgclpJ49x
-avp4Sgk3wLVue5iOKqlZL5fQIjckQEVR8vieKnZgGx6amVS9a5gB0GbAhkD06Y4p
-M3O6VQ==
-----END CERTIFICATE-----
--- a/static/vpnclient.ovpn
+++ b/static/vpnclient.ovpn
@@ -1,140 +0,0 @@
-verb 3
-nobind
-dev tun
-client
-remote 24.4.129.175 1194 udp
-fast-io
-compress lzo
-auth-nocache
-remote-cert-tls server
-<tls-crypt>
-----BEGIN OpenVPN Static key V1-----
-d188baecfc63820df3a11c50aa887c4e7236ff8021049038aec03f4f2a46376b
-aee8d80d06dbd812b84962937bed7003fdf64c264e9b7423925dbce4dd38b4e0
-a3bdfe6e656550a63430338c0dd4bcd4c694221c7561fa9e6da3efd0334a57ee
-5926acc05f768339b4712bf005d7eeb27f2da8dc8f4861b718b6683eb42869c0
-e11a1ac6c36daea5c79d7e08830de1c6f0a55207bb39e9c0420db34b3a631975
-5cfcef448f6664fde5d40e31e381503a6a724eebd7cfd76fe6d7108edc83b5ab
-ea1e66af70837d15a9d8ba58c82018b4cd669deb2323ba60d7c7ea8a398483aa
-2dec8aa6890dc2f60ff5be1a5c2a6a2fe95efa27f75c38735335e7f6f39b256e
-----END OpenVPN Static key V1-----
-</tls-crypt>
-<ca>
-----BEGIN CERTIFICATE-----
-MIIFNTCCAx2gAwIBAgIJAJBOAeknPeLqMA0GCSqGSIb3DQEBCwUAMDExCzAJBgNV
-BAYTAkdCMQ8wDQYDVQQIDAZMb25kb24xETAPBgNVBAoMCFdXVyBMdGQuMB4XDTE5
-MDEwNjA3NTIxOVoXDTI5MDEwMzA3NTIxOVowMTELMAkGA1UEBhMCR0IxDzANBgNV
-BAgMBkxvbmRvbjERMA8GA1UECgwIV1dXIEx0ZC4wggIiMA0GCSqGSIb3DQEBAQUA
-A4ICDwAwggIKAoICAQDpq5CFMT1VWb2MeaHXi4FpCLDXwnzaS+3qGCa3COdNg2BD
-tkQOPJNTgVhGn5XcfSDZnnVpXXrPDAqEDCUVVZj/2Mup/LseNr4miY+QcojyRETh
-Ecq0FVqgRvW2zRxqWxEPpLyzZGOcwAcW2jGO8XWPsqN4wAWO3WlpYT3unVK833Cx
-0wkdFIPbkEE1xKaJiskNYGgDuHu4tzGhOHSKOMzo7HvMaYsNgNChx/x4HYyTgkCf
-4r5zo4+CnOqRZ58STiV9AsOgg6mR8m6h/E9GNWpU0VWKm8hnklP+TiMnW/AKby0B
-hUKXFJMyhrNBOQXyj1LTqM5Q97+SNstOfqutKZgdD8mZcL4ec+DzelCH4Gyc15Yx
-gII/z3YwBUw/SGh+diCtWY2eJAHDkDFMGgiidVSzeKjRAgCDGi5+SYymLzLDyQey
-BpgbxunC2zHKsEhH1ZfOxyEOsW7UzgN6axQQ5DdzzKc1ke6OBl0YD1pRsoWEXudi
-b8LlNNI4oOaMiW3gsptJGPCOvXBrMm7wuzLrXMMRD0bh969KBJ7YQjUVkrAOsGTq
-DnqoXILa0ljsdazxe2Xk8GqrGAQ1XIvO7elbUlV/0nlAj4nzzx1m8f0n9nZ1aEZe
-Mv46+si6K/DgdUyGqcxOw6iZ00Fj6ha4yx7HJjZHHwFBXqJEPtdXYJKYa1AmHwID
-AQABo1AwTjAdBgNVHQ4EFgQUeBJp2fBea1UzyKirF1VYDsYddiwwHwYDVR0jBBgw
-FoAUeBJp2fBea1UzyKirF1VYDsYddiwwDAYDVR0TBAUwAwEB/zANBgkqhkiG9w0B
-AQsFAAOCAgEAfN7KjdqydCuSH6GCziu7Jc+SxLKLqMQc27/SBRWJ54JwDsfRglH3
-zze9j0f/auLKNirbxQG8/CeJO7BtxsPHk2NfKnXUMyIfRH4jlSmuuy0YLH1N3F19
-5GKGyt/ufc4a19l7M8ZseFMee8GXn6uHpVtN88GMKqQOu0AGnxv379ulI/RQ7iC2
-wkFpkT8Anzwd+jxMi5iNYbsHGd1uCyzY1bbNORY/fdX7A27xNjLe2cJc68OUOJQe
-XyfVlH2JyY+qEAXmv5gABafLFOsGmGHaQxZj4+zIdvDX6DGVIKCK7eixwVnKDwHm
-b9yF4ivMWk5gaY0sjezD7bnN2vAN1zXvpmmSu2tc/kOzGXZKoGEUn/4j+tWvvhPn
-wrTonT9soGmm7/LVyG/z950lylZV3XRw/0ZVQeCtQj+b+SjozNjTutzgWiAJ4njm
-Jyaqrj6vHB6vOPySk6AYyu1qTaJsniHR62Hv6WG/eZQalcXJZ8BuwAgdpcgPwdVU
-4IaKyiCjHg7dnrAwPURHfmlvosq+J+8PdD0O2L2aYUUtBS2TezgedSLXBYD4xZFa
-85zsZMlEurHM9o93vfjihyMxUla46o6uNyl32ebaPvLxEj/MyGOwkzAWa0qxy74J
-aQjWl+dWivXNFfE/yD/7yVF+X9YdlSFGCRyIfkUwy9hxLqkUdXeFgwE=
-----END CERTIFICATE-----
-</ca>
-<cert>
-----BEGIN CERTIFICATE-----
-MIIE5TCCAs2gAwIBAgICEAEwDQYJKoZIhvcNAQELBQAwMTELMAkGA1UEBhMCR0Ix
-DzANBgNVBAgMBkxvbmRvbjERMA8GA1UECgwIV1dXIEx0ZC4wHhcNMTkwMTA2MDc1
-MzQ5WhcNMjkwMTAzMDc1MzQ5WjAUMRIwEAYDVQQDDAl2cG5jbGllbnQwggIiMA0G
-CSqGSIb3DQEBAQUAA4ICDwAwggIKAoICAQCmpYnmgBet9aZInpN9DRi6EHjye699
-GFCVuBc4zw8PoZieC/jt/hQ8jrwQC8KVU7g2nuAZNfX1wXJ0hLDKAZWSJhvlAeN1
-/oA9oe/kikatUcijJnfipFJlhPJ8kru+UXH8ypwHoxbxd/2u+KDTjg3dJqNPGUak
-2KlSxHDbS0OWEhBIdn4A2iD7HpDPBO805KfAWVtfQ6Pvy5XHHNm6s796x91hes3Q
-ONY1TLKE+tHFMoTXn3C0a4/DU35Kj1JrZDdER+DrNmxGhInq2CXEGsyw8MwXYHLS
-1H8jvuX8u7yS2QX2cjwzc4PcJFPjEPTtUqPb1Ob+xOYb0bGOEE64+7xNeA+Mk5XL
-i08DJEFAU0hSCnHtz1y/JrXScKaHgtVzDm2TsXal9jO7ikVi3zhC6EcO0T4PFBX0
-lACXUDpXv50WI3ftoIACtO+paxM+wuXCt9ZhC6BZbRBs9EtVUdBDXAsSp+yhb9Jr
-GZzY+GIfmiPY027Cd0FYumspCHgBvUM4D9rCDVSiwr1DqNHqzlrjbwW3mCHLtmjE
-qU5Jg9DBUB3J01AERUgBYE8O8BkmSkuKi0mwgi9LdQ4SbizqVreMip+kXFAwRMg8
-Pw1h/cDUd9G1ZM+bZzHrjp4rdDHK8NAhDgJvxhGuhuwVFpA/LhvJ4tJBWjhJn7j5
-IEHpNWA1xNp1IwIDAQABoyQwIjALBgNVHQ8EBAMCB4AwEwYDVR0lBAwwCgYIKwYB
-BQUHAwIwDQYJKoZIhvcNAQELBQADggIBAFcqTpXcEe83shOnp+nOvGscMT0PwSNo
-ojy5xR9UhHs3ijyJ3DeaCO4xh7V8PTzCTpg3NOs8+19/nAhSr+QBWQKwrQhQ7Uub
-zv8AMXJ4tU1ZAyx0lX8FzUe/GsI8muqosK8F09jnTgGk05yCca9kDVzffGk1mivx
-d6ANRdUkprZV1VPA/eKXBQYstbeYBitPql5anmh54fEvXt7S1SdPATXI/eTzaxtP
-2KyPl7OZDA+mvS0qPFcY+MB2fjjdoyl74BShCJyI5sBCfN6WY6hNQ7meVWa6bCLQ
-EgvrZqh0lkWhy3mKcTL8eZJeF2SoYHQCSY75gQM0gdCODHTvkJJknLVmtzHHTJL3
-gVbdqFo/OiGGpD4XKpChNv//1kkRrwPBG4YDXu2/vsPoZKRVgjpQbNol+z+Ee1y3
-MqGz09aGxC98KUuxrUYwT8fbVVyLm0Fu2O8u+Qz3s6dPkWqD94YGxh2pq9SD5aTl
-/92LaIyqfMlWXj38yDUxjsENsTDtsSrx5cw6+BpB+VMmSuXlIYgE4khiEnYzCXbj
-5rduGmz4t4rhZZaa3n3L+G0sCQUqmnYNAjEMYcKIZvkTI3GoW3s0FROeUL1zLir2
-mdvWmQHTq39p/iBWmMTP/YofQPv8P1TWbKWaKalAf4+fLXHiTL7KHFw4YXXKE1iA
-GI6Ngos0UzHR
-----END CERTIFICATE-----
-</cert>
-<key>
-----BEGIN ENCRYPTED PRIVATE KEY-----
-MIIJljBABgkqhkiG9w0BBQ0wMzAbBgkqhkiG9w0BBQwwDgQITIzpduKGeaYCAggA
-MBQGCCqGSIb3DQMHBAjjV2SrWFzjJQSCCVBzWLc0eFF9x2OmoOddeslRnNd0DdiZ
-eOjsK93BkhlSpBJVLJY7x6DD3JJokgDCFl/sWHjU4zn4C5UogaqhrAIYeFpjx4w4
-4adU8bb3K4WOHAbQk5f+76HWr8LlCb4Ws2x0e8OlVwRBNJKAfumAoODE2ZZ0qitt
-5FeBix1XecaSpFl8J4BoytFD1R1Pf1KHL/iZ9Vh0SGGqE+ElDPOd8+PfsIKy09ZD
-/kMiiItnCF3NwG+s69GJJbGhIPdaas/yqVjtKOdQ8y7VBbrERzZ1mOVWN8zQj4cl
-5whPtMBgSYgkM9UcDNaQbqn/q4yXjPF+mWuZ6EyD4yNwpbroHEna5SAbaMiIHUah
-gXbSDrbFAMPqbpnpG5pUO9xM8YI8VYhlJU6MtzTjNwkmwyCzhV9WAEMBrMO3ZQGc
-FmG9HchAdKO75K7bHaaAZbBvt2LMMg+cvMSFiojKvOKXrC3ntQzHCrGL0IPEmdYK
-r1SIBD42zjsYPEY/MD0aV6eP/8DHbTMhF1oaQxxGLotv2+yVzoI+MTBaBFanQ+41
-7LBSB8oK8uq11vVz4LIp4xC+uyH+qoKE59mPG6QoRKC5GSjIj6J9hbDOS0DcHILg
-S0ebY2s2pSpVfcLZJslzSoI3ArNCs0fdkhZ2wD9/kX9BRAtZQSOsAcNVXc2sQsCj
-zeIZ9V7HlNGdZnG2CAPI5RWe8RSzz1T+IF9rUUD43Hi0csQ6y3IFQEXJtmXArVXo
-F6WoxqpXF3IdvLcnTDX1CK+h+QztSRysiRvWCPbISv84BIlx6OEVu+c0D82D+AUz
-Wf6DRsXIzqFKly/MZNsYG7Sx0t1eHaKaw+SCsWLRdiFsdmL+LUAcqVsJNCshKp6H
-Qlg3w0g9eU//qt2HnE0dx597PeSnyjRYSswt2R5dSaDh6x9KUeXc+kcTJTwxQ66c
-gSopFZyoGOxHLGoCBZV1qGGKbUVnbX8hy3eunVRNVsgOBFhmMYy1kaWajOGIfVSz
-jErclJpCJjuJDnK5L9ipLpQtb2VbVIgVbzwQ+p6AGBU39YO7R/ql4/DUyvo35mMx
-X9tr8uGYRWxkbBJSKZ6FNG0jUI++7goT66vMWb9Sn3Xsczj1J9INMeY4OGwXGZ+3
-VZrVsPMed0IJ4NIYJ0FRVhv7Y04aexJmvHqLUeRdJLk4l9kJNHoKJoleT8IUhThl
-nqP76jFabL3jX9fUpDxPNNoMiz+en4L8bX3dnLlvo8xeLnUaqT63Y+CgRVyVZJSI
-7PUZwGBWFHpuboLTYMgaQK6+UOp/rqtDFAkBXRD0ncSL2KYcy6I5IN9YDcYvTqvU
-N5TsVjftGKCCZAFyCkVqVjfV3uKJAiK4LHJe9J94Aq3lWeaw//gg1UjWrXCRwKuZ
-hO0kOEN7tw7YxOSOEzyQ3+j3TnWrToF/9QrCfY/+tOvwAVmLTD1e/dNTCt/SboaY
-2FGSI0TmPRSewxCT2L9hBgM1wtDdgSofVAwxW9qK+/0JPZm+C6gGc2ipZNdDH4uN
-+5j0zKZ98u7w6xRW23wCV+cnJ7IvtpXZvChFUnwnq9WanJakr/zsNsuhGpVOnD46
-QOZzO3U8VwXwK1yMas0oN7XSTwf0vGZYBaCtKkF7PrLlVeOYaj50jQNXUvfLt+bQ
-c64apwATE3JK2FcaV32m8UPz3bF04uuMIxBldH3Mvyp0X+MXaLERiuefUZwwppp5
-yFYuy+z03asOYeQrG8LSsTGNOgJXPu1Or32GMHlil1s74uOodA5T8XHEmX4Fxuik
-ok2itZL3yo/Sl73AC9yeSr7R9+Hf7SUTBt8AVeNPhmNDSi1AbyYhT1y0G7Dqwxvb
-oG5ZQyKPlquf8a5Xzodq6lPdXJwi8ZLmAuBelAg6A4MJZzMMrhOzQXyiMHVNtwdk
-c3LzES9bKEWgJR1CGR2RWrxUfqV6Y+uC+r5nPU/DZSOjJ0u7kWWvdQiySakxhQhO
-qkT/+PeYcspzB2juDA5kq9s5Votyl05nHoM7L3UdRxzA4IhKawL9lRWu0Q/0Gn4C
-axG2hUi61rxsx/epruIBz/01dlxw6xUpZBviArGxx/Z+QD0e5tKiuNCAOGQ5mm61
-UtopO6vBq6oS4O/xO/xHhYlcLKJ2D3C8v1JTDQXIQ1OY6IaqIXRiaB4/bz9eCByV
-8tEVW8/zs03M9zcxNuRL4tuBU0yhUCTCtGgQGUMnOYl4d1ZvZvUHb3oJqEI3AVOJ
-/tTVni6P3V4TdaF17EyndLZbIz9mRp3Tai4lsZXbRpevKzQfkFvg3vRZKj5Fymp6
-4SypXMKvgAQ/R6m9T8L0/rT90jf4GHhKhbSYXkJmAZQ/yz9eyjP9SFgHk3P0/z0x
-pa8oc50PAumDpz73pLFwYGffZb7yAAb+Uv4bjBgw+UytIWsbWJBmGAZpVOxKCFEX
-NGcflCfANB7FGv72a6fm3cf4IeqS3KGGQalIzOwwtIDWe2SVkp2LSx5JeFFCIjDP
-dKynm1tczPfoL/tUzcoRqI61zVpb3pAzKmrsWnSgA5Zl+LPZq23g5QRjCNeu1xkh
-GXMeXvQ0Q1VfLG4iw7j6zx52qiFy0HTQ8FK6cSA/nJN6/fE/2p+buAKxP5qjsFCP
-+/QQB681rfKGrQV1yh8TKuJ04h5gdxF4sC6cliHdw8daA2y4rQorBjM7F1EF8VmZ
-NcC1cEclv/E3QwOkBJsaom2rw7LkeOHLjqorGAf5eazO2AFZXVVG5yWrNyZWnaYf
-LYrXCk/4yLSexVEgiC81uSQL2uhvkatrUdDi4zV9mMrHKR10w8LVEuXSkS8IK3h5
-ln+HDc+rqUZG0ChHaF/GJ5VpQ6BLcMYNaoc75AuYU2rlSvMWnaR9UdiNVx3nrxld
-/SvNn8K+lFiKCr0J0DiVDztCpGOq4k2JSlCr+C+YxvipRr+VZOzpxx4RvkRFKAq6
-ix0demDcAk+YB6OZP3JAEy/yoiK/f61KiRpv0VVnHRFKyBv6MIyZmXkn5SesXF5C
-aBAV1zRdnV4EHXZy3qKIdvDP/5qp/6WcNI4edkAwr9bl+BqMe+0dy6QcsU9dLeQa
-OcpDZqHOxCXYTtiSIVM5WvSfPI5j6OdXCsrDU0VZOiiKegnGKNhz8Hn1aLZpGmoU
-TkqhRGpXchHSXNsGwT9AWlSJCnEF1dT0OOJzYbIbcwLa3WcKXHADpgfLJJ/KXHDJ
-buf/Epyjpi6dgg==
-----END ENCRYPTED PRIVATE KEY-----
-</key>
--- a/tags/index.html
+++ b/tags/index.html
@@ -0,0 +1,7 @@
+<!doctype html><html lang=en><head><title>Tags · Eric X. Liu's Personal Page</title><meta charset=utf-8><meta name=viewport content="width=device-width,initial-scale=1"><meta name=color-scheme content="light dark"><meta name=author content="Eric X. Liu"><meta name=description content="Eric X. Liu - Software & Performance Engineer at Google. Sharing insights about software engineering, performance optimization, tech industry experiences, mountain biking adventures, Jeep overlanding, and outdoor activities."><meta name=keywords content="software engineer,performance engineering,Google engineer,tech blog,software development,performance optimization,Eric Liu,engineering blog,mountain biking,Jeep enthusiast,overlanding,camping,outdoor adventures"><meta name=twitter:card content="summary"><meta name=twitter:title content="Tags"><meta name=twitter:description content="Eric X. Liu - Software & Performance Engineer at Google. Sharing insights about software engineering, performance optimization, tech industry experiences, mountain biking adventures, Jeep overlanding, and outdoor activities."><meta property="og:url" content="/tags/"><meta property="og:site_name" content="Eric X. Liu's Personal Page"><meta property="og:title" content="Tags"><meta property="og:description" content="Eric X. Liu - Software & Performance Engineer at Google. Sharing insights about software engineering, performance optimization, tech industry experiences, mountain biking adventures, Jeep overlanding, and outdoor activities."><meta property="og:locale" content="en"><meta property="og:type" content="website"><link rel=canonical href=/tags/><link rel=preload href=/fonts/fa-brands-400.woff2 as=font type=font/woff2 crossorigin><link rel=preload href=/fonts/fa-regular-400.woff2 as=font type=font/woff2 crossorigin><link rel=preload href=/fonts/fa-solid-900.woff2 as=font type=font/woff2 crossorigin><link rel=stylesheet href=/css/coder.min.6445a802b9389c9660e1b07b724dcf5718b1065ed2d71b4eeaf981cc7cc5fc46.css integrity="sha256-ZEWoArk4nJZg4bB7ck3PVxixBl7S1xtO6vmBzHzF/EY=" crossorigin=anonymous media=screen><link rel=stylesheet href=/css/coder-dark.min.a00e6364bacbc8266ad1cc81230774a1397198f8cfb7bcba29b7d6fcb54ce57f.css integrity="sha256-oA5jZLrLyCZq0cyBIwd0oTlxmPjPt7y6KbfW/LVM5X8=" crossorigin=anonymous media=screen><link rel=icon type=image/svg+xml href=/images/favicon.svg sizes=any><link rel=icon type=image/png href=/images/favicon-32x32.png sizes=32x32><link rel=icon type=image/png href=/images/favicon-16x16.png sizes=16x16><link rel=apple-touch-icon href=/images/apple-touch-icon.png><link rel=apple-touch-icon sizes=180x180 href=/images/apple-touch-icon.png><link rel=manifest href=/site.webmanifest><link rel=mask-icon href=/images/safari-pinned-tab.svg color=#5bbad5><link rel=alternate type=application/rss+xml href=/tags/index.xml title="Eric X. Liu's Personal Page"></head><body class="preload-transitions colorscheme-auto"><div class=float-container><a id=dark-mode-toggle class=colorscheme-toggle><i class="fa-solid fa-adjust fa-fw" aria-hidden=true></i></a></div><main class=wrapper><nav class=navigation><section class=container><a class=navigation-title href=/>Eric X. Liu's Personal Page
+</a><input type=checkbox id=menu-toggle>
+<label class="menu-button float-right" for=menu-toggle><i class="fa-solid fa-bars fa-fw" aria-hidden=true></i></label><ul class=navigation-list><li class=navigation-item><a class=navigation-link href=/posts/>Posts</a></li><li class=navigation-item><a class=navigation-link href=https://chat.ericxliu.me>Chat</a></li><li class=navigation-item><a class=navigation-link href=https://git.ericxliu.me/user/oauth2/Authenitk>Git</a></li><li class=navigation-item><a class=navigation-link href=https://coder.ericxliu.me/api/v2/users/oidc/callback>Coder</a></li><li class=navigation-item><a class=navigation-link href=/>|</a></li><li class=navigation-item><a class=navigation-link href=https://sso.ericxliu.me>Sign in</a></li></ul></section></nav><div class=content><section class="container taxonomy"><header><h1 class=title><a class=title-link href=/tags/>Tags</a></h1></header><ul></ul></section></div><footer class=footer><section class=container>©
+2016 -
+2025
+Eric X. Liu
+<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/69c2890">[69c2890]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script><script defer src=https://static.cloudflareinsights.com/beacon.min.js data-cf-beacon='{"token": "987638e636ce4dbb932d038af74c17d1"}'></script></body></html>
--- a/tags/index.xml
+++ b/tags/index.xml
@@ -0,0 +1 @@
+<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Tags on Eric X. Liu's Personal Page</title><link>/tags/</link><description>Recent content in Tags on Eric X. Liu's Personal Page</description><generator>Hugo</generator><language>en</language><atom:link href="/tags/index.xml" rel="self" type="application/rss+xml"/></channel></rss>
--- a/themes/hugo-coder
+++ b/themes/hugo-coder
Author	SHA1	Message	Date
eric	dbe2d5d1b0	deploy: `69c2890aad`	2025-08-20 06:29:20 +00:00
eric	2aadf95801	deploy: `16732daa57`	2025-08-20 06:24:28 +00:00
eric	ea9c28dce4	deploy: `3ee20f1118`	2025-08-20 06:05:30 +00:00
eric	1be19a7328	deploy: `ba596e75db`	2025-08-20 06:04:04 +00:00
eric	073fbfe081	deploy: `ed94cec2ad`	2025-08-20 04:50:28 +00:00
eric	ed03d0a873	deploy: `0945204182`	2025-08-20 04:34:01 +00:00
eric	798e6c7d75	deploy: `c06e978bd6`	2025-08-20 04:16:56 +00:00
eric	dff213a604	deploy: `4eabec8fbb`	2025-08-18 21:13:31 +00:00
eric	238fcb29b4	deploy: `849b7ef8d1`	2025-08-18 21:00:59 +00:00
eric	dc3978a294	deploy: `f0a4c1ed6e`	2025-08-16 23:38:42 +00:00
eric	6dfed70e80	deploy: `79c5380e51`	2025-08-16 23:03:09 +00:00
eric	596dc4948b	deploy: `beb7fbb468`	2025-08-16 22:58:02 +00:00
eric	cb921d30e0	deploy: `e33aef8159`	2025-08-16 21:43:36 +00:00
eric	50e9f52f56	deploy: `466bfee32b`	2025-08-16 21:14:31 +00:00
eric	b4e2b7f818	deploy: `bcab4969f4`	2025-08-16 21:08:19 +00:00
eric	c2b8a4f233	deploy: `a7f1af6c7f`	2025-08-16 20:59:44 +00:00
eric	8d18da2143	deploy: `fe52a73b7e`	2025-08-16 20:49:22 +00:00
eric	34ee48a56c	deploy: `8518b064dd`	2025-08-16 20:48:34 +00:00
eric	df3c006010	deploy: `2a5ee6dd11`	2025-08-14 06:51:16 +00:00
eric	c8813b97f3	deploy: `c9ed800d9f`	2025-08-09 04:06:01 +00:00
eric	52a6e87d0d	deploy: `c25cd8921e`	2025-08-04 03:59:59 +00:00
eric	5e1e4efc08	deploy: `de23f04b62`	2025-08-03 15:42:44 +00:00
eric	f50ba780e1	deploy: `abed5c59ab`	2025-08-03 15:38:00 +00:00
eric	a9192dd7da	deploy: `eba296fed3`	2025-08-03 06:15:25 +00:00
eric	a50fee0dcf	deploy: `5706ff7d48`	2025-08-03 05:45:33 +00:00
eric	9454edc7ed	deploy: `936d04687f`	2025-08-03 05:33:26 +00:00
eric	9efdd85826	deploy: `b4cc0126df`	2025-08-03 05:05:55 +00:00
eric	95df119b6d	deploy: `fd19c595b6`	2025-08-03 04:41:31 +00:00
eric	a6a4ee4adb	deploy: `f90b459eda`	2025-08-03 03:45:17 +00:00
eric	a977deebd1	deploy: `9c5d4a2102`	2025-08-03 03:30:50 +00:00
eric	8c3be83b91	deploy: `23b9adc43a`	2025-08-03 03:20:19 +00:00
eric	76c539f415	deploy: `38bbe8cbae`	2025-08-03 03:18:44 +00:00
eric	c1be16072c	deploy: `e15a7ffcef`	2025-08-03 03:17:32 +00:00
eric	11b8ac016c	deploy: `84b3c2016e`	2025-08-03 03:16:29 +00:00
eric	d03a2c49dd	deploy: `ec6f60a996`	2025-08-03 03:11:20 +00:00
eric	0ae24eb647	deploy: `38bbe8cbae`	2025-08-03 03:10:48 +00:00
eric	ce7b6b17b2	deploy: `bfc38f9b43`	2025-08-03 03:10:08 +00:00
eric	ef26adac81	deploy: `b213faf97d`	2025-08-03 03:09:27 +00:00
eric	fb47a09d9b	deploy: `015353bb06`	2025-08-03 03:06:54 +00:00
eric	b98d88fd0f	deploy: `afebdd8473`	2025-08-03 03:03:30 +00:00
eric	144a1b1692	deploy: `2a163cf7fe`	2025-08-03 02:54:11 +00:00
eric	df6ffb4bc0	deploy: `cbccd87961`	2025-08-03 02:47:13 +00:00
eric	219a24e3a5	deploy: `73f53ff6b9`	2025-08-03 02:41:00 +00:00
eric	335ed1d107	deploy: `38bbe8cbae`	2025-08-03 02:27:29 +00:00
eric	8f3c545991	deploy: `b6192ca3ca`	2025-08-03 02:20:56 +00:00
eric	e8ae2242e3	deploy: `a3ccac4cd2`	2025-08-02 22:50:04 +00:00
eric	d801fe9307	deploy: `88cbb7efd5`	2025-08-02 22:46:40 +00:00
eric	20c1888f78	deploy: `e6c8ccf0f8`	2025-08-02 22:45:16 +00:00
eric	9603629d20	deploy: `4996ae35fb`	2024-11-11 03:37:49 +00:00
eric	e60475c8ac	deploy: `b146b015d0`	2024-11-11 03:33:08 +00:00
eric	e83c0477c7	deploy: `53dd77638e`	2024-11-11 03:31:49 +00:00
eric	e86aa5f8cb	deploy: `f09740923b`	2024-11-11 02:37:20 +00:00
eric	8832dff8d6	deploy: `dfe3a7b0c8`	2024-09-16 21:18:23 +00:00
eric	ebed172a21	deploy: `9881f6e8bc`	2024-09-16 20:56:40 +00:00
eric	e7fda8a866	deploy: `d96148c395`	2024-09-16 20:09:44 +00:00
eric	a147bbd8c4	deploy: `248138f653`	2024-09-16 20:07:43 +00:00
eric	38518686d9	deploy: `8727ffa33f`	2024-09-16 18:38:19 +00:00
eric	38703cd607	deploy: `39e6a27cee`	2024-09-16 17:31:46 +00:00
eric	22b4234f06	deploy: `a8a360bf89`	2024-07-03 21:31:02 +00:00
eric	ef9bc708e1	deploy: `c5f30e590d`	2024-06-13 02:33:15 +00:00
eric Liu	61a3e5a38d	Drone build `1bdc086`	2024-06-12 06:10:05 +00:00
eric Liu	303714c386	Drone build `c68d13c`	2024-06-12 05:10:20 +00:00
eric Liu	203b36bc6c	Drone build `1f5623a`	2024-06-12 04:10:13 +00:00
eric Liu	085d1dd3f7	Drone build `d7a045f`	2024-06-11 19:10:11 +00:00
eric Liu	1f3238519a	Drone build `a6d01f0`	2024-04-25 22:26:25 +00:00
eric Liu	7ab352cdde	Drone build `d76023d`	2024-01-01 00:19:05 +00:00
eric Liu	e5c7ad2ee3	Drone build `d76023d`	2023-12-08 19:19:07 +00:00
eric Liu	b14698604d	Drone build `291f598`	2023-09-24 05:26:19 +00:00
eric Liu	482899015a	Drone build `4b89cb4`	2023-09-21 04:57:16 +00:00
eric Liu	396b46d31e	Drone build `6172792`	2023-09-19 04:23:58 +00:00
eric Liu	1d53e2965c	Drone build `64ed4f0`	2023-09-19 04:14:41 +00:00
eric Liu	77bd58c48f	Drone build `41a3fc0`	2023-09-19 04:06:51 +00:00
eric Liu	4e79964a24	Drone build `0fc636c`	2023-09-19 03:55:31 +00:00
eric Liu	7667b0ebf3	Drone build `607be11`	2023-09-19 02:53:59 +00:00
eric Liu	a9765b4d5b	Drone build `35a02ff`	2023-09-19 02:44:28 +00:00
eric Liu	d5b6868b70	Drone build `7c0df2c`	2023-09-19 02:20:42 +00:00
eric Liu	bd7fe9345f	Drone build `25b1f1f`	2023-09-18 19:20:23 +00:00
eric Liu	f20a18d653	Drone build `9b45a24`	2023-09-18 05:05:17 +00:00
eric Liu	c6d8e2aae6	Drone build `02d83d4`	2023-09-18 04:55:09 +00:00
eric Liu	685d7272e1	Drone build `23430e6`	2023-09-18 04:20:30 +00:00
eric Liu	2f5387a7a3	Drone build `9a38aab`	2023-09-18 04:10:58 +00:00
eric Liu	4b1dd1a9bf	Drone build `bac464a`	2023-09-18 03:56:53 +00:00
eric Liu	c05622c64f	Drone build `e715591`	2023-09-18 03:49:10 +00:00
eric Liu	b562560bbb	Drone build `cb2ac7b`	2023-09-11 03:19:13 +00:00
				`@@ -0,0 +1 @@`
				`<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Categories on Eric X. Liu's Personal Page</title><link>/categories/</link><description>Recent content in Categories on Eric X. Liu's Personal Page</description><generator>Hugo</generator><language>en</language><atom:link href="/categories/index.xml" rel="self" type="application/rss+xml"/></channel></rss>`
				`@@ -0,0 +1 @@`
				const body=document.body,darkModeToggle=document.getElementById("dark-mode-toggle"),darkModeMediaQuery=window.matchMedia("(prefers-color-scheme: dark)");localStorage.getItem("colorscheme")?setTheme(localStorage.getItem("colorscheme")):setTheme(body.classList.contains("colorscheme-light")\|\|body.classList.contains("colorscheme-dark")?body.classList.contains("colorscheme-dark")?"dark":"light":darkModeMediaQuery.matches?"dark":"light"),darkModeToggle&&darkModeToggle.addEventListener("click",()=>{let e=body.classList.contains("colorscheme-dark")?"light":"dark";setTheme(e),rememberTheme(e)}),darkModeMediaQuery.addListener(e=>{setTheme(e.matches?"dark":"light")}),document.addEventListener("DOMContentLoaded",function(){let e=document.querySelector(".preload-transitions");e.classList.remove("preload-transitions")});function setTheme(e){body.classList.remove("colorscheme-auto");let n=e==="dark"?"light":"dark";body.classList.remove("colorscheme-"+n),body.classList.add("colorscheme-"+e),document.documentElement.style["color-scheme"]=e;function t(e){return new Promise(t=>{if(document.querySelector(e))return t(document.querySelector(e));const n=new MutationObserver(s=>{document.querySelector(e)&&(t(document.querySelector(e)),n.disconnect())});n.observe(document.body,{childList:!0,subtree:!0})})}if(e==="dark"){const e={type:"set-theme",theme:"github-dark"};t(".utterances-frame").then(t=>{t.contentWindow.postMessage(e,"https://utteranc.es")})}else{const e={type:"set-theme",theme:"github-light"};t(".utterances-frame").then(t=>{t.contentWindow.postMessage(e,"https://utteranc.es")})}function s(e){const t=document.querySelector("iframe.giscus-frame");if(!t)return;t.contentWindow.postMessage({giscus:e},"https://giscus.app")}s({setConfig:{theme:e}});const o=new Event("themeChanged");document.dispatchEvent(o)}function rememberTheme(e){localStorage.setItem("colorscheme",e)}
				`@@ -0,0 +1 @@`
				`<!doctype html><html lang=en><head><title>/posts/</title><link rel=canonical href=/posts/><meta name=robots content="noindex"><meta charset=utf-8><meta http-equiv=refresh content="0; url=/posts/"></head></html>`
				`@@ -0,0 +1 @@`
				<?xml version="1.0" encoding="utf-8" standalone="yes"?><urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9" xmlns:xhtml="http://www.w3.org/1999/xhtml"><url><loc>/</loc><lastmod>2025-08-20T06:28:39+00:00</lastmod><changefreq>weekly</changefreq><priority>0.5</priority></url><url><loc>/posts/</loc><lastmod>2025-08-20T06:28:39+00:00</lastmod><changefreq>weekly</changefreq><priority>0.5</priority></url><url><loc>/posts/quantization-in-llms/</loc><lastmod>2025-08-20T06:02:35+00:00</lastmod><changefreq>weekly</changefreq><priority>0.5</priority></url><url><loc>/posts/breville-barista-pro-maintenance/</loc><lastmod>2025-08-20T06:04:36+00:00</lastmod><changefreq>weekly</changefreq><priority>0.5</priority></url><url><loc>/posts/secure-boot-dkms-and-mok-on-proxmox-debian/</loc><lastmod>2025-08-14T06:50:22+00:00</lastmod><changefreq>weekly</changefreq><priority>0.5</priority></url><url><loc>/posts/how-rvq-teaches-llms-to-see-and-hear/</loc><lastmod>2025-08-08T17:36:52+00:00</lastmod><changefreq>weekly</changefreq><priority>0.5</priority></url><url><loc>/posts/supabase-deep-dive/</loc><lastmod>2025-08-04T03:59:37+00:00</lastmod><changefreq>weekly</changefreq><priority>0.5</priority></url><url><loc>/posts/ppo-for-language-models/</loc><lastmod>2025-08-20T06:04:36+00:00</lastmod><changefreq>weekly</changefreq><priority>0.5</priority></url><url><loc>/posts/mixture-of-experts-moe-models-challenges-solutions-in-practice/</loc><lastmod>2025-08-03T06:02:48+00:00</lastmod><changefreq>weekly</changefreq><priority>0.5</priority></url><url><loc>/posts/t5-the-transformer-that-zigged-when-others-zagged-an-architectural-deep-dive/</loc><lastmod>2025-08-03T03:41:10+00:00</lastmod><changefreq>weekly</changefreq><priority>0.5</priority></url><url><loc>/posts/espresso-theory-application-a-guide-for-the-breville-barista-pro/</loc><lastmod>2025-08-03T04:20:20+00:00</lastmod><changefreq>weekly</changefreq><priority>0.5</priority></url><url><loc>/posts/transformer-s-core-mechanics/</loc><lastmod>2025-08-20T06:28:39+00:00</lastmod><changefreq>weekly</changefreq><priority>0.5</priority></url><url><loc>/posts/useful/</loc><lastmod>2025-08-03T08:37:28-07:00</lastmod><changefreq>weekly</changefreq><priority>0.5</priority></url><url><loc>/about/</loc><lastmod>2020-06-16T23:30:17-07:00</lastmod><changefreq>weekly</changefreq><priority>0.5</priority></url><url><loc>/categories/</loc><changefreq>weekly</changefreq><priority>0.5</priority></url><url><loc>/tags/</loc><changefreq>weekly</changefreq><priority>0.5</priority></url></urlset>