deploy: fd19c595b6
This commit is contained in:
eric
2
404.html
2
404.html
@@ -4,4 +4,4 @@
|
||||
2016 -
|
||||
2025
|
||||
Eric X. Liu
|
||||
<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/f90b459">[f90b459]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script></body></html>
|
||||
<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/fd19c59">[fd19c59]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script></body></html>
|
||||
@@ -4,4 +4,4 @@
|
||||
2016 -
|
||||
2025
|
||||
Eric X. Liu
|
||||
<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/f90b459">[f90b459]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script></body></html>
|
||||
<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/fd19c59">[fd19c59]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script></body></html>
|
||||
@@ -4,4 +4,4 @@
|
||||
2016 -
|
||||
2025
|
||||
Eric X. Liu
|
||||
<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/f90b459">[f90b459]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script></body></html>
|
||||
<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/fd19c59">[fd19c59]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script></body></html>
|
||||
@@ -4,4 +4,4 @@
|
||||
2016 -
|
||||
2025
|
||||
Eric X. Liu
|
||||
<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/f90b459">[f90b459]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script></body></html>
|
||||
<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/fd19c59">[fd19c59]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script></body></html>
|
||||
@@ -1,4 +1,4 @@
|
||||
<?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>Sun, 03 Aug 2025 03:41:10 +0000</lastBuildDate><atom:link href="/index.xml" rel="self" type="application/rss+xml"/><item><title>A Deep Dive into PPO for Language Models</title><link>/posts/a-deep-dive-into-ppo-for-language-models/</link><pubDate>Sat, 02 Aug 2025 00:00:00 +0000</pubDate><guid>/posts/a-deep-dive-into-ppo-for-language-models/</guid><description><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>
|
||||
<?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>Sun, 03 Aug 2025 04:20:20 +0000</lastBuildDate><atom:link href="/index.xml" rel="self" type="application/rss+xml"/><item><title>A Deep Dive into PPO for Language Models</title><link>/posts/a-deep-dive-into-ppo-for-language-models/</link><pubDate>Sat, 02 Aug 2025 00:00:00 +0000</pubDate><guid>/posts/a-deep-dive-into-ppo-for-language-models/</guid><description><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></description></item><item><title>Mixture-of-Experts (MoE) Models Challenges & 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><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
|
||||
@@ -8,8 +8,9 @@
|
||||
</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></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><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></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><ul>
|
||||
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></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><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></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><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></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><ul>
|
||||
<li><a href="https://ericxliu.me/rootCA.pem" class="external-link" target="_blank" rel="noopener">rootCA.pem</a></li>
|
||||
<li><a href="https://ericxliu.me/vpnclient.ovpn" class="external-link" target="_blank" rel="noopener">vpnclient.ovpn</a></li>
|
||||
</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>
|
||||
@@ -23,4 +23,4 @@ where <code>δ_t = r_t + γV(s_{t+1}) - V(s_t)</code></p><ul><li><strong>γ (gam
|
||||
2016 -
|
||||
2025
|
||||
Eric X. Liu
|
||||
<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/f90b459">[f90b459]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script></body></html>
|
||||
<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/fd19c59">[fd19c59]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script></body></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=fediverse:creator content><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.60f552a2c0452fcc0254c54f21c3e0728460c1ae85f97a9c35833a222ef8b884.css integrity="sha256-YPVSosBFL8wCVMVPIcPgcoRgwa6F+XqcNYM6Ii74uIQ=" 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=https://rss.ericxliu.me/oauth2/oidc/redirect>RSS</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’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’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 “headspace,” 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 “Work”:</strong> A higher dose means more coffee mass, requiring more “work” (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’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’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’s the <strong>result</strong> of how much resistance your coffee puck provides against the machine’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 (<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’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’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’t flow evenly. Instead, it “breaks” 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 “Grind Amount” 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? (<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’ll be pulling consistently delicious espresso shots from your Breville Barista Pro in no time.</p></div><footer></footer></article></section></div><footer class=footer><section class=container>©
|
||||
2016 -
|
||||
2025
|
||||
Eric X. Liu
|
||||
<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/fd19c59">[fd19c59]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script></body></html>
|
||||
@@ -3,9 +3,10 @@
|
||||
<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=https://rss.ericxliu.me/oauth2/oidc/redirect>RSS</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 2, 2025</span>
|
||||
<a class=title href=/posts/a-deep-dive-into-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>October 26, 2020</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>October 26, 2020</span>
|
||||
<a class=title href=/posts/useful/>Some useful files</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/f90b459">[f90b459]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script></body></html>
|
||||
<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/fd19c59">[fd19c59]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script></body></html>
|
||||
@@ -1,4 +1,4 @@
|
||||
<?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>Sun, 03 Aug 2025 03:41:10 +0000</lastBuildDate><atom:link href="/posts/index.xml" rel="self" type="application/rss+xml"/><item><title>A Deep Dive into PPO for Language Models</title><link>/posts/a-deep-dive-into-ppo-for-language-models/</link><pubDate>Sat, 02 Aug 2025 00:00:00 +0000</pubDate><guid>/posts/a-deep-dive-into-ppo-for-language-models/</guid><description><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>
|
||||
<?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>Sun, 03 Aug 2025 04:20:20 +0000</lastBuildDate><atom:link href="/posts/index.xml" rel="self" type="application/rss+xml"/><item><title>A Deep Dive into PPO for Language Models</title><link>/posts/a-deep-dive-into-ppo-for-language-models/</link><pubDate>Sat, 02 Aug 2025 00:00:00 +0000</pubDate><guid>/posts/a-deep-dive-into-ppo-for-language-models/</guid><description><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></description></item><item><title>Mixture-of-Experts (MoE) Models Challenges & 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><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
|
||||
@@ -8,8 +8,9 @@
|
||||
</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></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><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></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><ul>
|
||||
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></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><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></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><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></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><ul>
|
||||
<li><a href="https://ericxliu.me/rootCA.pem" class="external-link" target="_blank" rel="noopener">rootCA.pem</a></li>
|
||||
<li><a href="https://ericxliu.me/vpnclient.ovpn" class="external-link" target="_blank" rel="noopener">vpnclient.ovpn</a></li>
|
||||
</ul></description></item></channel></rss>
|
||||
@@ -7,9 +7,9 @@
|
||||
|
||||
|
||||
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 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=fediverse:creator content><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="2025-08-03T03:28:39+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.60f552a2c0452fcc0254c54f21c3e0728460c1ae85f97a9c35833a222ef8b884.css integrity="sha256-YPVSosBFL8wCVMVPIcPgcoRgwa6F+XqcNYM6Ii74uIQ=" 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
|
||||
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 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=fediverse:creator content><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="2025-08-03T03:49:59+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.60f552a2c0452fcc0254c54f21c3e0728460c1ae85f97a9c35833a222ef8b884.css integrity="sha256-YPVSosBFL8wCVMVPIcPgcoRgwa6F+XqcNYM6Ii74uIQ=" 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=https://rss.ericxliu.me/oauth2/oidc/redirect>RSS</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
|
||||
@@ -17,25 +17,25 @@ Many routing mechanisms, especially “Top-K routing,” involve a discr
|
||||
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 “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.</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 “Top-K routing,” 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><p>$$
|
||||
Many routing mechanisms, especially “Top-K routing,” involve a discrete, hard selection process. A common function is <code>KeepTopK(v, k)</code>, which selects the top <code>k</code> scoring elements from a vector <code>v</code> and sets others to (-\infty) or (0).</p><p>[
|
||||
KeepTopK(v, k)_i = \begin{cases} v_i & \text{if } v_i \text{ is in the top } k \text{ elements of } v \ -\infty & \text{otherwise.} \end{cases}
|
||||
$$</p><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’s parameters (e.g., $W_g$) using standard gradient descent.</p><p><strong>Solutions (Stochastic Approximations):</strong>
|
||||
]</p><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’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 $H(x)_i = (x \cdot W_g)<em>i + \text{StandardNormal}() \cdot \text{Softplus}((x \cdot W</em>{noise})_i)$ 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>
|
||||
The expert score (H(x)_i = (x \cdot W_g)<em>i + \text{StandardNormal}() \cdot \text{Softplus}((x \cdot W</em>{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’s parameters are updated based on the “reward” (e.g., the model’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)
|
||||
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’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><p>$$ \text{loss}<em>{\text{auxiliary}} = \alpha \cdot N \cdot \sum</em>{i=1}^{N} f_i \cdot P_i $$</p><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 “more frequent use = stronger downweighting,” 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’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’s not a pre-defined value.</li></ul><p><strong>Limitation (“Second Best” Scenario):</strong>
|
||||
Even with this loss, an expert can remain imbalanced if it’s consistently the “second best” 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 “soft” routing or “softmax after TopK” approaches can be more effective for truly even distribution.</p><h3 id=3-challenge-overfitting-during-fine-tuning>3. Challenge: Overfitting during Fine-tuning
|
||||
An auxiliary loss is added to the total model loss during training to encourage more even expert usage.</p><p>(( \text{loss}<em>{\text{auxiliary}} = \alpha \cdot N \cdot \sum</em>{i=1}^{N} f_i \cdot P_i ))</p><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 “more frequent use = stronger downweighting,” 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’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’s not a pre-defined value.</li></ul><p><strong>Limitation (“Second Best” Scenario):</strong>
|
||||
Even with this loss, an expert can remain imbalanced if it’s consistently the “second best” 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 “soft” routing or “softmax after TopK” 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’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’s parameters during fine-tuning, specifically the large expert weights.</li><li>Instead, only the “non-MoE” 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’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>
|
||||
@@ -44,4 +44,4 @@ The <strong>Top-K routing</strong> mechanism, as illustrated in the provided ima
|
||||
2016 -
|
||||
2025
|
||||
Eric X. Liu
|
||||
<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/f90b459">[f90b459]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script></body></html>
|
||||
<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/fd19c59">[fd19c59]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script></body></html>
|
||||
@@ -30,4 +30,4 @@ But to truly understand the field, we must look at the pivotal models that explo
|
||||
2016 -
|
||||
2025
|
||||
Eric X. Liu
|
||||
<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/f90b459">[f90b459]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script></body></html>
|
||||
<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/fd19c59">[fd19c59]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script></body></html>
|
||||
@@ -10,4 +10,4 @@ One-minute read</span></div></div></header><div class=post-content><ul><li><a hr
|
||||
2016 -
|
||||
2025
|
||||
Eric X. Liu
|
||||
<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/f90b459">[f90b459]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script></body></html>
|
||||
<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/fd19c59">[fd19c59]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script></body></html>
|
||||
@@ -1 +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>/posts/a-deep-dive-into-ppo-for-language-models/</loc><lastmod>2025-08-03T03:28:39+00:00</lastmod><changefreq>weekly</changefreq><priority>0.5</priority></url><url><loc>/</loc><lastmod>2025-08-03T03:41:10+00:00</lastmod><changefreq>weekly</changefreq><priority>0.5</priority></url><url><loc>/posts/</loc><lastmod>2025-08-03T03:41:10+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>2025-08-03T03:28:39+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/useful/</loc><lastmod>2020-10-26T04:47:36+00:00</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>
|
||||
<?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>/posts/a-deep-dive-into-ppo-for-language-models/</loc><lastmod>2025-08-03T03:28:39+00:00</lastmod><changefreq>weekly</changefreq><priority>0.5</priority></url><url><loc>/</loc><lastmod>2025-08-03T04:20:20+00:00</lastmod><changefreq>weekly</changefreq><priority>0.5</priority></url><url><loc>/posts/</loc><lastmod>2025-08-03T04:20:20+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>2025-08-03T03:49:59+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/useful/</loc><lastmod>2020-10-26T04:47:36+00:00</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>
|
||||
@@ -4,4 +4,4 @@
|
||||
2016 -
|
||||
2025
|
||||
Eric X. Liu
|
||||
<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/f90b459">[f90b459]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script></body></html>
|
||||
<a href="https://git.ericxliu.me/eric/ericxliu-me/commit/fd19c59">[fd19c59]</a></section></footer></main><script src=/js/coder.min.6ae284be93d2d19dad1f02b0039508d9aab3180a12a06dcc71b0b0ef7825a317.js integrity="sha256-auKEvpPS0Z2tHwKwA5UI2aqzGAoSoG3McbCw73gloxc="></script></body></html>
|
||||
Reference in New Issue
Block a user