Spaces:
Running
Running
File size: 6,319 Bytes
85928ca f98af2f 85928ca f98af2f 85928ca | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 | <!doctype html>
<html lang="en">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width,initial-scale=1">
<title>PyC Platform</title>
<link rel="stylesheet" href="./assets/css/site.css">
</head>
<body>
<main class="container">
<header class="site-header">
<div class="top-bar">
<p class="eyebrow">PyC Compiler Runtime</p>
<button id="theme-toggle" type="button" class="theme-toggle" aria-label="Toggle color theme">Switch Theme</button>
</div>
<nav class="site-nav" aria-label="Primary">
<a class="active" href="./index.html">Main</a>
<a href="./results-insights.html">Results Insights</a>
<a href="./kernel-results.html">Latest Kernel Results</a>
<a href="./inference/index.html">Inference Portal</a>
</nav>
</header>
<section class="hero">
<p class="hero-kicker">Production compiler rails for ML systems</p>
<h1>Distributed training with deterministic controls, measurable throughput, and visible bottleneck telemetry.</h1>
<p class="hero-copy">
PyC packages compiler-next contracts, runtime fallbacks, and benchmark publication into a single operational loop.
CPU orchestration and GPU execution are intentionally split, then rejoined through instrumentation so performance changes are explainable.
</p>
<div class="hero-cta">
<a id="release-link" class="cta primary" href="#" target="_blank" rel="noopener noreferrer">Latest release</a>
<a class="cta" href="./results-insights.html">Open results hub</a>
<a class="cta" href="./inference/index.html">Open inference portal</a>
</div>
<p id="status" class="status">Loading release metadata...</p>
</section>
<section class="flow">
<h2>Operational Flow</h2>
<p class="section-text">
Runtime stages are coordinated as a conveyor: host-side preprocessing, pinned-memory transfer, GPU compute, communication sync,
then telemetry publication. This keeps throughput high while preserving deterministic rollback behavior.
</p>
<div class="pill-row">
<span class="pill">Pinned host memory staging</span>
<span class="pill">Async H2D dispatch</span>
<span class="pill">NCCL synchronized gradient flow</span>
<span class="pill">Artifact publication pipeline</span>
</div>
</section>
<section class="flow">
<h2>Install and Validate</h2>
<pre><code>cmake -S . -B build -D PYC_BUILD_COMPILER_NEXT=ON -D PYC_BUILD_COMPILER_NEXT_TESTS=ON
cmake --build build --parallel
ctest --test-dir build -C Release --output-on-failure
./build/pyc</code></pre>
<p class="section-text">Binary downloads:</p>
<ul class="downloads">
<li>Linux: <a id="download-linux" href="https://github.com/DarkStarStrix/PyC/releases/latest/download/pyc-linux-x86_64.tar.gz">pyc-linux-x86_64.tar.gz</a></li>
<li>macOS: <a id="download-macos" href="https://github.com/DarkStarStrix/PyC/releases/latest/download/pyc-macos-arm64.tar.gz">pyc-macos-arm64.tar.gz</a></li>
<li>Windows: <a id="download-windows" href="https://github.com/DarkStarStrix/PyC/releases/latest/download/pyc-windows-x86_64.zip">pyc-windows-x86_64.zip</a></li>
</ul>
</section>
<section class="flow">
<h2>Latest Distributed Evidence</h2>
<p id="distributed-status" class="status">Loading distributed training insights...</p>
<div class="kpi-grid" id="dist-kpi-grid"></div>
<div class="chart-grid chart-grid-wide">
<figure>
<img id="latest-dist-summary-main" src="" alt="Latest distributed run summary chart" loading="lazy">
<figcaption>Latest campaign summary</figcaption>
</figure>
<figure>
<img id="latest-dist-throughput-main" src="" alt="Latest distributed throughput chart" loading="lazy">
<figcaption>Distributed throughput comparison</figcaption>
</figure>
<figure>
<img id="latest-dist-pipeline-main" src="" alt="Latest distributed pipeline breakdown chart" loading="lazy">
<figcaption>Latest pipeline breakdown</figcaption>
</figure>
</div>
<p class="section-text">
Published artifacts:
<a href="./results/manifest.json" target="_blank" rel="noopener noreferrer">manifest.json</a>
|
<a href="./results/latest-summary.json" target="_blank" rel="noopener noreferrer">latest-summary.json</a>
|
<a href="./results/distributed-latest.json" target="_blank" rel="noopener noreferrer">distributed-latest.json</a>
|
<a href="./kernel-results.html">latest kernel suite</a>
</p>
</section>
<section class="flow">
<h2>Compiler Adapter Baseline</h2>
<p id="results-status" class="status">Loading benchmark publication data...</p>
<h3>CPU Adapter Summary</h3>
<table>
<thead>
<tr>
<th>Adapter</th>
<th>Mode</th>
<th>Mean (ms)</th>
<th>P50 (ms)</th>
<th>P95 (ms)</th>
<th>Throughput</th>
</tr>
</thead>
<tbody id="cpu-results-body"></tbody>
</table>
<h3>GPU Adapter Summary</h3>
<table>
<thead>
<tr>
<th>Adapter</th>
<th>Mode</th>
<th>Mean (ms)</th>
<th>P50 (ms)</th>
<th>P95 (ms)</th>
<th>Throughput</th>
</tr>
</thead>
<tbody id="gpu-results-body"></tbody>
</table>
<div class="chart-grid">
<figure>
<img id="latest-cpu-svg" src="./results/artifacts/latest/latest_cpu.svg" alt="Latest CPU benchmark chart" loading="lazy">
<figcaption>CPU baseline snapshot</figcaption>
</figure>
<figure>
<img id="latest-gpu-svg" src="./results/artifacts/latest/latest_gpu.svg" alt="Latest GPU benchmark chart" loading="lazy">
<figcaption>GPU baseline snapshot</figcaption>
</figure>
</div>
</section>
<section class="flow">
<h2>Release Assets</h2>
<ul id="asset-list" class="asset-list"></ul>
</section>
</main>
<script type="module" src="./assets/js/main.js"></script>
</body>
</html>
|