mirror of
https://github.com/harvard-edge/cs249r_book.git
synced 2026-03-09 07:15:51 -05:00
34b350dbec
- Rename collabs/ directory to labs/ - Rename workflow files: collabs-publish-*.yml → labs-publish-*.yml - Update all paths in workflows, Quarto configs, and navbars - Update cross-references in book, kits, and dev-landing - Update site-url to mlsysbook.ai/labs/ Note: GitHub repo variable DEV_COLLABS_PATH needs to be renamed to DEV_LABS_PATH in repository settings.
109 lines
3.4 KiB
Plaintext
109 lines
3.4 KiB
Plaintext
---
|
|
pagetitle: "Co-Labs"
|
|
title: ""
|
|
page-layout: full
|
|
toc: false
|
|
format:
|
|
html:
|
|
include-in-header:
|
|
text: ""
|
|
---
|
|
|
|
```{=html}
|
|
<div class="hero-section">
|
|
<div class="hero-content">
|
|
<span class="coming-badge">Coming 2026</span>
|
|
<h1>Co-Labs</h1>
|
|
<p class="hero-subtitle">See ML Systems in Action</p>
|
|
<p class="hero-tagline">Watch quantization compress models. Measure memory hierarchies. Profile gradient flow.</p>
|
|
</div>
|
|
</div>
|
|
|
|
<div class="content-section">
|
|
|
|
<h2>Why Another Set of Notebooks?</h2>
|
|
|
|
<p>There are many excellent Colab notebooks for ML. Most demonstrate <em>algorithms</em>. Few help you understand <em>systems</em>.</p>
|
|
|
|
<p>When you quantize a model from FP32 to INT8, what actually happens to the weights? When you increase batch size, where does memory go? When you add a layer, how does gradient flow change?</p>
|
|
|
|
<p>Co-Labs are designed to answer these questions. Each notebook is <strong>systematically aligned with textbook chapters</strong>, letting you experiment with the exact system concepts you just read about. The goal is not to teach you how to use PyTorch. It's to show you <em>why PyTorch works the way it does</em>.</p>
|
|
|
|
<p><em>— Vijay</em></p>
|
|
|
|
<h2>The Systems Learning Path</h2>
|
|
|
|
<p>Co-Labs fit between conceptual understanding and building from scratch.</p>
|
|
|
|
<div class="journey-cards">
|
|
<div class="journey-card">
|
|
<div class="journey-icon">📖</div>
|
|
<h3>Understand</h3>
|
|
<p>Learn system design principles: memory, compute, parallelism, efficiency</p>
|
|
<a href="../book/">Textbook →</a>
|
|
</div>
|
|
<div class="journey-card active">
|
|
<div class="journey-icon">🔬</div>
|
|
<h3>Experiment</h3>
|
|
<p>Measure tradeoffs, profile bottlenecks, see system decisions ripple through models</p>
|
|
<span class="current">Co-Labs</span>
|
|
</div>
|
|
<div class="journey-card">
|
|
<div class="journey-icon">🔥</div>
|
|
<h3>Build</h3>
|
|
<p>Implement tensors, autograd, and training loops from scratch</p>
|
|
<a href="../tinytorch/">TinyTorch →</a>
|
|
</div>
|
|
</div>
|
|
|
|
<h2>What You'll Explore</h2>
|
|
|
|
<p>Each Co-Lab maps directly to textbook chapters, focusing on the <em>systems</em> perspective:</p>
|
|
|
|
<div class="topic-grid">
|
|
<div class="topic-card">
|
|
<h4>Memory Systems</h4>
|
|
<ul>
|
|
<li>Batch size vs memory footprint</li>
|
|
<li>Activation checkpointing tradeoffs</li>
|
|
<li>Cache hierarchy effects on training</li>
|
|
</ul>
|
|
</div>
|
|
<div class="topic-card">
|
|
<h4>Numerical Representation</h4>
|
|
<ul>
|
|
<li>FP32 → FP16 → INT8 → INT4</li>
|
|
<li>Quantization error propagation</li>
|
|
<li>Mixed precision training dynamics</li>
|
|
</ul>
|
|
</div>
|
|
<div class="topic-card">
|
|
<h4>Compute Efficiency</h4>
|
|
<ul>
|
|
<li>Pruning and sparsity patterns</li>
|
|
<li>Knowledge distillation mechanics</li>
|
|
<li>Operator fusion benefits</li>
|
|
</ul>
|
|
</div>
|
|
<div class="topic-card">
|
|
<h4>Deployment Tradeoffs</h4>
|
|
<ul>
|
|
<li>Latency vs throughput curves</li>
|
|
<li>Batching strategy impact</li>
|
|
<li>Hardware utilization profiling</li>
|
|
</ul>
|
|
</div>
|
|
</div>
|
|
|
|
<h2>Help Shape This</h2>
|
|
|
|
<p>I'm still figuring out what makes the most sense. If you have ideas for experiments that would help you understand ML systems better:</p>
|
|
|
|
<div class="cta-section">
|
|
<a href="https://github.com/harvard-edge/cs249r_book/discussions" class="cta-btn primary">Share Ideas</a>
|
|
<a href="#subscribe" class="cta-btn secondary">Get Updates</a>
|
|
</div>
|
|
|
|
</div>
|
|
```
|