chore: import upstream snapshot with attribution
This commit is contained in:
wehub-resource-sync
@@ -0,0 +1 @@
|
||||
nn.labml.ai
|
||||
@@ -0,0 +1,209 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content=""/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="configs.py"/>
|
||||
<meta name="twitter:description" content=""/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/rwkv/configs.html"/>
|
||||
<meta property="og:title" content="configs.py"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="configs.py"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="configs.py"/>
|
||||
<meta property="og:description" content=""/>
|
||||
|
||||
<title>configs.py</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/rwkv/configs.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="index.html">rwkv</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/rwkv/configs.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">1</span><span></span><span class="kn">from</span> <span class="nn">labml.configs</span> <span class="kn">import</span> <span class="n">BaseConfigs</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-1'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-1'>#</a>
|
||||
</div>
|
||||
<h2>Transformer Configurations</h2>
|
||||
<p>This defines configurations for a transformer. The configurations are calculate using option functions. These are lazy loaded and therefore only the necessary modules are calculated.</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">4</span><span class="k">class</span> <span class="nc">RWKVConfigs</span><span class="p">(</span><span class="n">BaseConfigs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-2'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-2'>#</a>
|
||||
</div>
|
||||
<p>Number of attention heads </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">14</span> <span class="n">n_heads</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">8</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-3'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-3'>#</a>
|
||||
</div>
|
||||
<p>Transformer embedding size </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">16</span> <span class="n">d_model</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">512</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-4'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-4'>#</a>
|
||||
</div>
|
||||
<p>Number of layers </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">18</span> <span class="n">n_layers</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">6</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-5'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-5'>#</a>
|
||||
</div>
|
||||
<p>Dropout probability </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">20</span> <span class="n">dropout</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">0.1</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-6'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-6'>#</a>
|
||||
</div>
|
||||
<p>Number of tokens in the source vocabulary (for token embeddings) </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">22</span> <span class="n">n_src_vocab</span><span class="p">:</span> <span class="nb">int</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-7'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-7'>#</a>
|
||||
</div>
|
||||
<p>Number of tokens in the target vocabulary (to generate logits for prediction) </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">24</span> <span class="n">n_tgt_vocab</span><span class="p">:</span> <span class="nb">int</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
@@ -0,0 +1,623 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content=""/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="experiment.py"/>
|
||||
<meta name="twitter:description" content=""/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/rwkv/experiment.html"/>
|
||||
<meta property="og:title" content="experiment.py"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="experiment.py"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="experiment.py"/>
|
||||
<meta property="og:description" content=""/>
|
||||
|
||||
<title>experiment.py</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/rwkv/experiment.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="index.html">rwkv</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/rwkv/experiment.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">1</span><span></span><span class="kn">import</span> <span class="nn">inspect</span>
|
||||
<span class="lineno">2</span><span class="kn">import</span> <span class="nn">math</span>
|
||||
<span class="lineno">3</span>
|
||||
<span class="lineno">4</span><span class="kn">import</span> <span class="nn">torch</span>
|
||||
<span class="lineno">5</span><span class="kn">import</span> <span class="nn">torch.nn</span> <span class="k">as</span> <span class="nn">nn</span>
|
||||
<span class="lineno">6</span><span class="kn">from</span> <span class="nn">labml_nn.rwkv.configs</span> <span class="kn">import</span> <span class="n">RWKVConfigs</span>
|
||||
<span class="lineno">7</span>
|
||||
<span class="lineno">8</span><span class="kn">from</span> <span class="nn">labml_nn.rwkv</span> <span class="kn">import</span> <span class="n">RWKV</span>
|
||||
<span class="lineno">9</span><span class="kn">from</span> <span class="nn">labml_nn.rwkv</span> <span class="kn">import</span> <span class="n">TimeMixing</span>
|
||||
<span class="lineno">10</span><span class="kn">from</span> <span class="nn">labml</span> <span class="kn">import</span> <span class="n">experiment</span>
|
||||
<span class="lineno">11</span><span class="kn">from</span> <span class="nn">labml.configs</span> <span class="kn">import</span> <span class="n">option</span>
|
||||
<span class="lineno">12</span><span class="kn">from</span> <span class="nn">labml_nn.experiments.nlp_autoregression</span> <span class="kn">import</span> <span class="n">NLPAutoRegressionConfigs</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-1'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-1'>#</a>
|
||||
</div>
|
||||
<h2>Configurations</h2>
|
||||
<p>This inherits from <a href="../../experiments/nlp_autoregression.html#NLPAutoRegressionConfigs"><code class="highlight"><span></span><span class="n">NLPAutoRegressionConfigs</span></code>
|
||||
</a></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">15</span><span class="k">class</span> <span class="nc">Configs</span><span class="p">(</span><span class="n">NLPAutoRegressionConfigs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-2'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-2'>#</a>
|
||||
</div>
|
||||
<p>RWKV model </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">24</span> <span class="n">model</span><span class="p">:</span> <span class="n">RWKV</span>
|
||||
<span class="lineno">25</span>
|
||||
<span class="lineno">26</span> <span class="n">rwkv</span><span class="p">:</span> <span class="n">RWKVConfigs</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-3'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-3'>#</a>
|
||||
</div>
|
||||
<p>number of warmup iterations </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">28</span> <span class="n">warmup_iters</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">2000</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-4'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-4'>#</a>
|
||||
</div>
|
||||
<p>total number of training iterations </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">30</span> <span class="n">max_iters</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">600000</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-5'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-5'>#</a>
|
||||
</div>
|
||||
<p>weight decay </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">32</span> <span class="n">weight_decay</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">1e-1</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-6'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-6'>#</a>
|
||||
</div>
|
||||
<p>Custom optimizer </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">34</span> <span class="n">beta1</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">0.9</span>
|
||||
<span class="lineno">35</span> <span class="n">beta2</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">0.95</span>
|
||||
<span class="lineno">36</span> <span class="n">optimizer</span> <span class="o">=</span> <span class="s1">'rwkv_optimizer'</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-7'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-7'>#</a>
|
||||
</div>
|
||||
<h3>RWKV configurations</h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">39</span><span class="nd">@option</span><span class="p">(</span><span class="n">Configs</span><span class="o">.</span><span class="n">rwkv</span><span class="p">,</span> <span class="s1">'RWKV'</span><span class="p">)</span>
|
||||
<span class="lineno">40</span><span class="k">def</span> <span class="nf">_rwkv_configs</span><span class="p">(</span><span class="n">c</span><span class="p">:</span> <span class="n">Configs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-8'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-8'>#</a>
|
||||
</div>
|
||||
<p>We use our <a href="../configs.html#RWKVConfigs">configurable RWKV implementation</a> </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">47</span> <span class="n">conf</span> <span class="o">=</span> <span class="n">RWKVConfigs</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-9'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-9'>#</a>
|
||||
</div>
|
||||
<p>Set the vocabulary sizes for embeddings and generating logits </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">49</span> <span class="n">conf</span><span class="o">.</span><span class="n">n_src_vocab</span> <span class="o">=</span> <span class="n">c</span><span class="o">.</span><span class="n">n_tokens</span>
|
||||
<span class="lineno">50</span> <span class="n">conf</span><span class="o">.</span><span class="n">n_tgt_vocab</span> <span class="o">=</span> <span class="n">c</span><span class="o">.</span><span class="n">n_tokens</span>
|
||||
<span class="lineno">51</span>
|
||||
<span class="lineno">52</span> <span class="k">return</span> <span class="n">conf</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-10'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-10'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">55</span><span class="k">def</span> <span class="nf">_init_weights</span><span class="p">(</span><span class="n">module</span><span class="p">,</span> <span class="n">rwkv</span><span class="p">:</span> <span class="n">RWKVConfigs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-11'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-11'>#</a>
|
||||
</div>
|
||||
<p>initialize Vector Parameters in TimeMixing </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">57</span> <span class="k">if</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">module</span><span class="p">,</span> <span class="n">TimeMixing</span><span class="p">):</span>
|
||||
<span class="lineno">58</span> <span class="n">layer_id</span> <span class="o">=</span> <span class="n">module</span><span class="o">.</span><span class="n">layer_id</span>
|
||||
<span class="lineno">59</span> <span class="n">n_layer</span> <span class="o">=</span> <span class="n">module</span><span class="o">.</span><span class="n">n_layer</span>
|
||||
<span class="lineno">60</span> <span class="n">n_embd</span> <span class="o">=</span> <span class="n">module</span><span class="o">.</span><span class="n">n_embd</span>
|
||||
<span class="lineno">61</span> <span class="n">attn_sz</span> <span class="o">=</span> <span class="n">n_embd</span>
|
||||
<span class="lineno">62</span>
|
||||
<span class="lineno">63</span> <span class="k">with</span> <span class="n">torch</span><span class="o">.</span><span class="n">no_grad</span><span class="p">():</span>
|
||||
<span class="lineno">64</span> <span class="n">ratio_0_to_1</span> <span class="o">=</span> <span class="n">layer_id</span> <span class="o">/</span> <span class="p">(</span><span class="n">n_layer</span> <span class="o">-</span> <span class="mi">1</span><span class="p">)</span> <span class="c1"># 0 to 1</span>
|
||||
<span class="lineno">65</span> <span class="n">ratio_1_to_almost0</span> <span class="o">=</span> <span class="mf">1.0</span> <span class="o">-</span> <span class="p">(</span><span class="n">layer_id</span> <span class="o">/</span> <span class="n">n_layer</span><span class="p">)</span> <span class="c1"># 1 to ~0</span>
|
||||
<span class="lineno">66</span> <span class="n">ddd</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">ones</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="n">n_embd</span><span class="p">)</span>
|
||||
<span class="lineno">67</span> <span class="k">for</span> <span class="n">i</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="n">n_embd</span><span class="p">):</span>
|
||||
<span class="lineno">68</span> <span class="n">ddd</span><span class="p">[</span><span class="mi">0</span><span class="p">,</span> <span class="mi">0</span><span class="p">,</span> <span class="n">i</span><span class="p">]</span> <span class="o">=</span> <span class="n">i</span> <span class="o">/</span> <span class="n">n_embd</span>
|
||||
<span class="lineno">69</span>
|
||||
<span class="lineno">70</span> <span class="n">decay_speed</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">ones</span><span class="p">(</span><span class="n">attn_sz</span><span class="p">)</span>
|
||||
<span class="lineno">71</span> <span class="k">for</span> <span class="n">h</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="n">attn_sz</span><span class="p">):</span>
|
||||
<span class="lineno">72</span> <span class="n">decay_speed</span><span class="p">[</span><span class="n">h</span><span class="p">]</span> <span class="o">=</span> <span class="o">-</span><span class="mi">5</span> <span class="o">+</span> <span class="mi">8</span> <span class="o">*</span> <span class="p">(</span><span class="n">h</span> <span class="o">/</span> <span class="p">(</span><span class="n">attn_sz</span> <span class="o">-</span> <span class="mi">1</span><span class="p">))</span> <span class="o">**</span> <span class="p">(</span><span class="mf">0.7</span> <span class="o">+</span> <span class="mf">1.3</span> <span class="o">*</span> <span class="n">ratio_0_to_1</span><span class="p">)</span>
|
||||
<span class="lineno">73</span> <span class="n">module</span><span class="o">.</span><span class="n">time_decay</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Parameter</span><span class="p">(</span><span class="n">decay_speed</span><span class="p">)</span>
|
||||
<span class="lineno">74</span>
|
||||
<span class="lineno">75</span> <span class="n">zigzag</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">tensor</span><span class="p">([(</span><span class="n">i</span> <span class="o">+</span> <span class="mi">1</span><span class="p">)</span> <span class="o">%</span> <span class="mi">3</span> <span class="o">-</span> <span class="mi">1</span> <span class="k">for</span> <span class="n">i</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="n">attn_sz</span><span class="p">)])</span> <span class="o">*</span> <span class="mf">0.5</span>
|
||||
<span class="lineno">76</span> <span class="n">module</span><span class="o">.</span><span class="n">time_first</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Parameter</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">ones</span><span class="p">(</span><span class="n">attn_sz</span><span class="p">)</span> <span class="o">*</span> <span class="n">math</span><span class="o">.</span><span class="n">log</span><span class="p">(</span><span class="mf">0.3</span><span class="p">)</span> <span class="o">+</span> <span class="n">zigzag</span><span class="p">)</span>
|
||||
<span class="lineno">77</span> <span class="n">module</span><span class="o">.</span><span class="n">time_mix_key</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Parameter</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">pow</span><span class="p">(</span><span class="n">ddd</span><span class="p">,</span> <span class="n">ratio_1_to_almost0</span><span class="p">))</span>
|
||||
<span class="lineno">78</span> <span class="n">module</span><span class="o">.</span><span class="n">time_mix_value</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Parameter</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">pow</span><span class="p">(</span><span class="n">ddd</span><span class="p">,</span> <span class="n">ratio_1_to_almost0</span><span class="p">)</span> <span class="o">+</span> <span class="mf">0.3</span> <span class="o">*</span> <span class="n">ratio_0_to_1</span><span class="p">)</span>
|
||||
<span class="lineno">79</span> <span class="n">module</span><span class="o">.</span><span class="n">time_mix_receptance</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Parameter</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">pow</span><span class="p">(</span><span class="n">ddd</span><span class="p">,</span> <span class="mf">0.5</span> <span class="o">*</span> <span class="n">ratio_1_to_almost0</span><span class="p">))</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-12'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-12'>#</a>
|
||||
</div>
|
||||
<p> Create RWKV model and initialize weights</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">82</span><span class="nd">@option</span><span class="p">(</span><span class="n">Configs</span><span class="o">.</span><span class="n">model</span><span class="p">)</span>
|
||||
<span class="lineno">83</span><span class="k">def</span> <span class="nf">_model</span><span class="p">(</span><span class="n">c</span><span class="p">:</span> <span class="n">Configs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-13'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-13'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">87</span> <span class="n">m</span> <span class="o">=</span> <span class="n">RWKV</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">rwkv</span><span class="p">)</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">device</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-14'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-14'>#</a>
|
||||
</div>
|
||||
<p>Apply custom weight initialization </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">90</span> <span class="n">m</span><span class="o">.</span><span class="n">apply</span><span class="p">(</span><span class="n">_init_weights</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">rwkv</span><span class="p">)</span>
|
||||
<span class="lineno">91</span>
|
||||
<span class="lineno">92</span> <span class="k">return</span> <span class="n">m</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-15'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-15'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">95</span><span class="nd">@option</span><span class="p">(</span><span class="n">NLPAutoRegressionConfigs</span><span class="o">.</span><span class="n">optimizer</span><span class="p">)</span>
|
||||
<span class="lineno">96</span><span class="k">def</span> <span class="nf">_configure_optimizers</span><span class="p">(</span><span class="n">c</span><span class="p">:</span> <span class="n">NLPAutoRegressionConfigs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-16'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-16'>#</a>
|
||||
</div>
|
||||
<p>start with all of the candidate parameters </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">98</span> <span class="n">param_dict</span> <span class="o">=</span> <span class="p">{</span><span class="n">pn</span><span class="p">:</span> <span class="n">p</span> <span class="k">for</span> <span class="n">pn</span><span class="p">,</span> <span class="n">p</span> <span class="ow">in</span> <span class="n">c</span><span class="o">.</span><span class="n">model</span><span class="o">.</span><span class="n">named_parameters</span><span class="p">()}</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-17'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-17'>#</a>
|
||||
</div>
|
||||
<p>filter out those that do not require grad </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">100</span> <span class="n">param_dict</span> <span class="o">=</span> <span class="p">{</span><span class="n">pn</span><span class="p">:</span> <span class="n">p</span> <span class="k">for</span> <span class="n">pn</span><span class="p">,</span> <span class="n">p</span> <span class="ow">in</span> <span class="n">param_dict</span><span class="o">.</span><span class="n">items</span><span class="p">()</span> <span class="k">if</span> <span class="n">p</span><span class="o">.</span><span class="n">requires_grad</span><span class="p">}</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-18'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-18'>#</a>
|
||||
</div>
|
||||
<p>create optim groups. Any parameters that is 2D will be weight decayed, otherwise no. i.e. all weight tensors in matmuls + embeddings decay, all biases and layernorms don't. </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">103</span> <span class="n">decay_params</span> <span class="o">=</span> <span class="p">[</span><span class="n">p</span> <span class="k">for</span> <span class="n">n</span><span class="p">,</span> <span class="n">p</span> <span class="ow">in</span> <span class="n">param_dict</span><span class="o">.</span><span class="n">items</span><span class="p">()</span> <span class="k">if</span> <span class="n">p</span><span class="o">.</span><span class="n">dim</span><span class="p">()</span> <span class="o">>=</span> <span class="mi">2</span><span class="p">]</span>
|
||||
<span class="lineno">104</span> <span class="n">nodecay_params</span> <span class="o">=</span> <span class="p">[</span><span class="n">p</span> <span class="k">for</span> <span class="n">n</span><span class="p">,</span> <span class="n">p</span> <span class="ow">in</span> <span class="n">param_dict</span><span class="o">.</span><span class="n">items</span><span class="p">()</span> <span class="k">if</span> <span class="n">p</span><span class="o">.</span><span class="n">dim</span><span class="p">()</span> <span class="o"><</span> <span class="mi">2</span><span class="p">]</span>
|
||||
<span class="lineno">105</span> <span class="n">optim_groups</span> <span class="o">=</span> <span class="p">[</span>
|
||||
<span class="lineno">106</span> <span class="p">{</span><span class="s1">'params'</span><span class="p">:</span> <span class="n">decay_params</span><span class="p">,</span> <span class="s1">'weight_decay'</span><span class="p">:</span> <span class="n">c</span><span class="o">.</span><span class="n">weight_decay</span><span class="p">},</span>
|
||||
<span class="lineno">107</span> <span class="p">{</span><span class="s1">'params'</span><span class="p">:</span> <span class="n">nodecay_params</span><span class="p">,</span> <span class="s1">'weight_decay'</span><span class="p">:</span> <span class="mf">0.0</span><span class="p">}</span>
|
||||
<span class="lineno">108</span> <span class="p">]</span>
|
||||
<span class="lineno">109</span> <span class="n">num_decay_params</span> <span class="o">=</span> <span class="nb">sum</span><span class="p">(</span><span class="n">p</span><span class="o">.</span><span class="n">numel</span><span class="p">()</span> <span class="k">for</span> <span class="n">p</span> <span class="ow">in</span> <span class="n">decay_params</span><span class="p">)</span>
|
||||
<span class="lineno">110</span> <span class="n">num_nodecay_params</span> <span class="o">=</span> <span class="nb">sum</span><span class="p">(</span><span class="n">p</span><span class="o">.</span><span class="n">numel</span><span class="p">()</span> <span class="k">for</span> <span class="n">p</span> <span class="ow">in</span> <span class="n">nodecay_params</span><span class="p">)</span>
|
||||
<span class="lineno">111</span> <span class="nb">print</span><span class="p">(</span><span class="sa">f</span><span class="s2">"num decayed parameter tensors: </span><span class="si">{</span><span class="nb">len</span><span class="p">(</span><span class="n">decay_params</span><span class="p">)</span><span class="si">}</span><span class="s2">, with </span><span class="si">{</span><span class="n">num_decay_params</span><span class="si">:</span><span class="s2">,</span><span class="si">}</span><span class="s2"> parameters"</span><span class="p">)</span>
|
||||
<span class="lineno">112</span> <span class="nb">print</span><span class="p">(</span><span class="sa">f</span><span class="s2">"num non-decayed parameter tensors: </span><span class="si">{</span><span class="nb">len</span><span class="p">(</span><span class="n">nodecay_params</span><span class="p">)</span><span class="si">}</span><span class="s2">, with </span><span class="si">{</span><span class="n">num_nodecay_params</span><span class="si">:</span><span class="s2">,</span><span class="si">}</span><span class="s2"> parameters"</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-19'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-19'>#</a>
|
||||
</div>
|
||||
<p>Create AdamW optimizer and use the fused version if it is available </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">114</span> <span class="n">fused_available</span> <span class="o">=</span> <span class="s1">'fused'</span> <span class="ow">in</span> <span class="n">inspect</span><span class="o">.</span><span class="n">signature</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">optim</span><span class="o">.</span><span class="n">AdamW</span><span class="p">)</span><span class="o">.</span><span class="n">parameters</span>
|
||||
<span class="lineno">115</span> <span class="n">use_fused</span> <span class="o">=</span> <span class="n">fused_available</span> <span class="ow">and</span> <span class="n">c</span><span class="o">.</span><span class="n">device_type</span> <span class="o">==</span> <span class="s1">'cuda'</span>
|
||||
<span class="lineno">116</span> <span class="n">extra_args</span> <span class="o">=</span> <span class="nb">dict</span><span class="p">(</span><span class="n">fused</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span> <span class="k">if</span> <span class="n">use_fused</span> <span class="k">else</span> <span class="nb">dict</span><span class="p">()</span>
|
||||
<span class="lineno">117</span> <span class="n">optimizer</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">optim</span><span class="o">.</span><span class="n">AdamW</span><span class="p">(</span><span class="n">optim_groups</span><span class="p">,</span> <span class="n">lr</span><span class="o">=</span><span class="n">c</span><span class="o">.</span><span class="n">learning_rate</span><span class="p">,</span> <span class="n">betas</span><span class="o">=</span><span class="n">c</span><span class="o">.</span><span class="n">betas</span><span class="p">,</span> <span class="o">**</span><span class="n">extra_args</span><span class="p">)</span>
|
||||
<span class="lineno">118</span> <span class="nb">print</span><span class="p">(</span><span class="sa">f</span><span class="s2">"using fused AdamW: </span><span class="si">{</span><span class="n">use_fused</span><span class="si">}</span><span class="s2">"</span><span class="p">)</span>
|
||||
<span class="lineno">119</span>
|
||||
<span class="lineno">120</span> <span class="k">return</span> <span class="n">optimizer</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-20'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-20'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">123</span><span class="k">def</span> <span class="nf">main</span><span class="p">():</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-21'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-21'>#</a>
|
||||
</div>
|
||||
<p>Create experiment </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">125</span> <span class="n">experiment</span><span class="o">.</span><span class="n">create</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s2">"RWKV"</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-22'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-22'>#</a>
|
||||
</div>
|
||||
<p>Create configs </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">127</span> <span class="n">conf</span> <span class="o">=</span> <span class="n">Configs</span><span class="p">()</span>
|
||||
<span class="lineno">128</span> <span class="nb">print</span><span class="p">(</span><span class="n">conf</span><span class="o">.</span><span class="n">model</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-23'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-23'>#</a>
|
||||
</div>
|
||||
<p>Override configurations </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">130</span> <span class="n">experiment</span><span class="o">.</span><span class="n">configs</span><span class="p">(</span><span class="n">conf</span><span class="p">,</span> <span class="p">{</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-24'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-24'>#</a>
|
||||
</div>
|
||||
<p>Use character level tokenizer </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">132</span> <span class="s1">'tokenizer'</span><span class="p">:</span> <span class="s1">'character'</span><span class="p">,</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-25'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-25'>#</a>
|
||||
</div>
|
||||
<p>Prompt separator is blank </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">134</span> <span class="s1">'prompt_separator'</span><span class="p">:</span> <span class="s1">''</span><span class="p">,</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-26'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-26'>#</a>
|
||||
</div>
|
||||
<p>Starting prompt for sampling </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">136</span> <span class="s1">'prompt'</span><span class="p">:</span> <span class="s1">'It is '</span><span class="p">,</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-27'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-27'>#</a>
|
||||
</div>
|
||||
<p>Use Tiny Shakespeare dataset </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">138</span> <span class="s1">'text'</span><span class="p">:</span> <span class="s1">'tiny_shakespeare'</span><span class="p">,</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-28'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-28'>#</a>
|
||||
</div>
|
||||
<p>Use a context size of <span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqa" style=""><span class="mord" style="">128</span></span></span></span></span></span> </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">141</span> <span class="s1">'seq_len'</span><span class="p">:</span> <span class="mi">128</span><span class="p">,</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-29'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-29'>#</a>
|
||||
</div>
|
||||
<p>Train for <span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord">32</span></span></span></span></span> epochs </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">143</span> <span class="s1">'epochs'</span><span class="p">:</span> <span class="mi">32</span><span class="p">,</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-30'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-30'>#</a>
|
||||
</div>
|
||||
<p>Batch size <span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqa" style=""><span class="mord" style="">128</span></span></span></span></span></span> </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">145</span> <span class="s1">'batch_size'</span><span class="p">:</span> <span class="mi">128</span><span class="p">,</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-31'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-31'>#</a>
|
||||
</div>
|
||||
<p>Switch between training and validation for <span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord">10</span></span></span></span></span> times per epoch </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">148</span> <span class="s1">'inner_iterations'</span><span class="p">:</span> <span class="mi">10</span><span class="p">,</span>
|
||||
<span class="lineno">149</span>
|
||||
<span class="lineno">150</span> <span class="s1">'rwkv.block_size'</span><span class="p">:</span> <span class="mi">1024</span><span class="p">,</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-32'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-32'>#</a>
|
||||
</div>
|
||||
<p>model </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">152</span> <span class="s1">'rwkv.n_layer'</span><span class="p">:</span> <span class="mi">12</span><span class="p">,</span>
|
||||
<span class="lineno">153</span> <span class="s1">'rwkv.n_heads'</span><span class="p">:</span> <span class="mi">12</span><span class="p">,</span>
|
||||
<span class="lineno">154</span> <span class="s1">'rwkv.n_embd'</span><span class="p">:</span> <span class="mi">768</span>
|
||||
<span class="lineno">155</span> <span class="p">})</span>
|
||||
<span class="lineno">156</span>
|
||||
<span class="lineno">157</span> <span class="nb">print</span><span class="p">(</span><span class="n">conf</span><span class="o">.</span><span class="n">model</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-33'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-33'>#</a>
|
||||
</div>
|
||||
<p>Set models for saving and loading </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">159</span> <span class="n">experiment</span><span class="o">.</span><span class="n">add_pytorch_models</span><span class="p">({</span><span class="s1">'model'</span><span class="p">:</span> <span class="n">conf</span><span class="o">.</span><span class="n">model</span><span class="p">})</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-34'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-34'>#</a>
|
||||
</div>
|
||||
<p>Start the experiment </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">162</span> <span class="k">with</span> <span class="n">experiment</span><span class="o">.</span><span class="n">start</span><span class="p">():</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-35'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-35'>#</a>
|
||||
</div>
|
||||
<p>Run training </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">164</span> <span class="n">conf</span><span class="o">.</span><span class="n">run</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-36'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-36'>#</a>
|
||||
</div>
|
||||
<p> </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">168</span><span class="k">if</span> <span class="vm">__name__</span> <span class="o">==</span> <span class="s1">'__main__'</span><span class="p">:</span>
|
||||
<span class="lineno">169</span> <span class="n">main</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
File diff suppressed because one or more lines are too long
@@ -0,0 +1,839 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content="在 Tiny Shakespeare 的 FFN 中使用自由贸易协定训练变压器。"/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="模糊平铺激活实验"/>
|
||||
<meta name="twitter:description" content="在 Tiny Shakespeare 的 FFN 中使用自由贸易协定训练变压器。"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/activations/fta/experiment.html"/>
|
||||
<meta property="og:title" content="模糊平铺激活实验"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="模糊平铺激活实验"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="模糊平铺激活实验"/>
|
||||
<meta property="og:description" content="在 Tiny Shakespeare 的 FFN 中使用自由贸易协定训练变压器。"/>
|
||||
|
||||
<title>模糊平铺激活实验</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/activations/fta/experiment.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="../index.html">activations</a>
|
||||
<a class="parent" href="index.html">fta</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/activations/fta/experiment.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1><a href="index.html">模糊拼贴激活</a>实验</h1>
|
||||
<p><a href="https://colab.research.google.com/github/labmlai/annotated_deep_learning_paper_implementations/blob/master/labml_nn/activations/fta/experiment.ipynb"><img alt="Open In Colab" src="https://colab.research.google.com/assets/colab-badge.svg"></a></p>
|
||||
<p>在这里，我们训练一台在<a href="../../transformers/feed_forward.html">前馈网络</a>中使用<a href="index.html">模糊切片激活</a>的变压器。我们将其用作语言模型，并在小莎士比亚数据集上对其进行训练以进行演示。</p>
|
||||
<p>但是，对于 FTA 来说，这可能不是理想的任务，我们认为 FTA 更适合对具有连续变量的数据进行建模。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">21</span><span></span><span class="kn">import</span> <span class="nn">copy</span>
|
||||
<span class="lineno">22</span>
|
||||
<span class="lineno">23</span><span class="kn">import</span> <span class="nn">torch</span>
|
||||
<span class="lineno">24</span><span class="kn">import</span> <span class="nn">torch.nn</span> <span class="k">as</span> <span class="nn">nn</span>
|
||||
<span class="lineno">25</span>
|
||||
<span class="lineno">26</span><span class="kn">from</span> <span class="nn">labml</span> <span class="kn">import</span> <span class="n">experiment</span>
|
||||
<span class="lineno">27</span><span class="kn">from</span> <span class="nn">labml.configs</span> <span class="kn">import</span> <span class="n">option</span>
|
||||
<span class="lineno">28</span><span class="kn">from</span> <span class="nn">labml_helpers.module</span> <span class="kn">import</span> <span class="n">Module</span>
|
||||
<span class="lineno">29</span><span class="kn">from</span> <span class="nn">labml_nn.activations.fta</span> <span class="kn">import</span> <span class="n">FTA</span>
|
||||
<span class="lineno">30</span><span class="kn">from</span> <span class="nn">labml_nn.experiments.nlp_autoregression</span> <span class="kn">import</span> <span class="n">NLPAutoRegressionConfigs</span>
|
||||
<span class="lineno">31</span><span class="kn">from</span> <span class="nn">labml_nn.transformers</span> <span class="kn">import</span> <span class="n">MultiHeadAttention</span><span class="p">,</span> <span class="n">TransformerLayer</span>
|
||||
<span class="lineno">32</span><span class="kn">from</span> <span class="nn">labml_nn.transformers.utils</span> <span class="kn">import</span> <span class="n">subsequent_mask</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-1'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-1'>#</a>
|
||||
</div>
|
||||
<h2>带有 F <a href="index.html">TA 激活功能的 FF</a> N 模块</h2>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">35</span><span class="k">class</span> <span class="nc">FeedForwardFTA</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-2'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-2'>#</a>
|
||||
</div>
|
||||
<ul><li><code class="highlight"><span></span><span class="n">d_model</span></code>
|
||||
是令牌嵌入中的要素数量</li>
|
||||
<li><code class="highlight"><span></span><span class="n">d_ff</span></code>
|
||||
是 FFN 隐藏层中的要素数量</li>
|
||||
<li><code class="highlight"><span></span><span class="n">activation</span></code>
|
||||
是 FTA 激活模块</li>
|
||||
<li><code class="highlight"><span></span><span class="n">dropout</span></code>
|
||||
是隐藏层的丢失概率</li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">40</span> <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">d_model</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">d_ff</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span>
|
||||
<span class="lineno">41</span> <span class="n">activation</span><span class="p">:</span> <span class="n">FTA</span><span class="p">,</span>
|
||||
<span class="lineno">42</span> <span class="n">dropout</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">0.1</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-3'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-3'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">49</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-4'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-4'>#</a>
|
||||
</div>
|
||||
<p>第一层按权重<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.83333em;vertical-align:-0.15em;"></span><span class="mord coloredeq eqe" style=""><span class="mord" style=""><span class="mord mathnormal" style="margin-right:0.13889em">W</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.30110799999999993em;"><span style="top:-2.5500000000000003em;margin-left:-0.13889em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style="">1</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span></span></span></span></span></span>和偏差进行参数化<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.84444em;vertical-align:-0.15em;"></span><span class="mord coloredeq eqf" style=""><span class="mord" style=""><span class="mord mathnormal" style="">b</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.30110799999999993em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style="">1</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">51</span> <span class="bp">self</span><span class="o">.</span><span class="n">layer1</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Linear</span><span class="p">(</span><span class="n">d_model</span><span class="p">,</span> <span class="n">d_ff</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-5'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-5'>#</a>
|
||||
</div>
|
||||
<p>第二层按权重<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.83333em;vertical-align:-0.15em;"></span><span class="mord coloredeq eqe" style=""><span class="mord" style=""><span class="mord mathnormal" style="margin-right:0.13889em">W</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.30110799999999993em;"><span style="top:-2.5500000000000003em;margin-left:-0.13889em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style="">1</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span></span></span></span></span></span>和偏差进行参数化<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.84444em;vertical-align:-0.15em;"></span><span class="mord coloredeq eqf" style=""><span class="mord" style=""><span class="mord mathnormal" style="">b</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.30110799999999993em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style="">1</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">53</span> <span class="bp">self</span><span class="o">.</span><span class="n">layer2</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Linear</span><span class="p">(</span><span class="n">d_ff</span> <span class="o">*</span> <span class="n">activation</span><span class="o">.</span><span class="n">expansion_factor</span><span class="p">,</span> <span class="n">d_model</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-6'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-6'>#</a>
|
||||
</div>
|
||||
<p>隐藏图层丢失</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">55</span> <span class="bp">self</span><span class="o">.</span><span class="n">dropout</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Dropout</span><span class="p">(</span><span class="n">dropout</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-7'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-7'>#</a>
|
||||
</div>
|
||||
<p>激活功能<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.8888799999999999em;vertical-align:-0.19444em;"></span><span class="mord coloredeq eqi" style=""><span class="mord mathnormal" style="margin-right:0.10764em">f</span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">57</span> <span class="bp">self</span><span class="o">.</span><span class="n">activation</span> <span class="o">=</span> <span class="n">activation</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-8'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-8'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">59</span> <span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-9'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-9'>#</a>
|
||||
</div>
|
||||
<p><span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:1em;vertical-align:-0.25em;"></span><span class="mord coloredeq eqi" style=""><span class="mord mathnormal" style="margin-right:0.10764em">f</span></span><span class="mopen">(</span><span class="mord mathnormal">x</span><span class="mord coloredeq eqe" style=""><span class="mord" style=""><span class="mord mathnormal" style="margin-right:0.13889em">W</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.30110799999999993em;"><span style="top:-2.5500000000000003em;margin-left:-0.13889em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style="">1</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span></span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin">+</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span></span><span class="base"><span class="strut" style="height:1em;vertical-align:-0.25em;"></span><span class="mord coloredeq eqf" style=""><span class="mord" style=""><span class="mord mathnormal" style="">b</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.30110799999999993em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style="">1</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span></span><span class="mclose">)</span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">61</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">activation</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">layer1</span><span class="p">(</span><span class="n">x</span><span class="p">))</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-10'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-10'>#</a>
|
||||
</div>
|
||||
<p>申请退学</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">63</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">dropout</span><span class="p">(</span><span class="n">x</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-11'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-11'>#</a>
|
||||
</div>
|
||||
<p></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">65</span> <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">layer2</span><span class="p">(</span><span class="n">x</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-12'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-12'>#</a>
|
||||
</div>
|
||||
<h2>自回归模型</h2>
|
||||
<p>这是一个自回归变压器模型，它使用前馈网络和（模糊平铺激活）（index.html）。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">68</span><span class="k">class</span> <span class="nc">AutoregressiveTransformer</span><span class="p">(</span><span class="n">Module</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-13'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-13'>#</a>
|
||||
</div>
|
||||
<ul><li><code class="highlight"><span></span><span class="n">n_tokens</span></code>
|
||||
是词汇表中代币的数量</li>
|
||||
<li><code class="highlight"><span></span><span class="n">d_model</span></code>
|
||||
是嵌入的大小</li>
|
||||
<li><code class="highlight"><span></span><span class="n">n_layers</span></code>
|
||||
是变压器层的数量</li>
|
||||
<li><code class="highlight"><span></span><span class="n">layer</span></code>
|
||||
是层。我们在变压器上使用这个<code class="highlight"><span></span><span class="n">n_layers</span></code>
|
||||
副本。</li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">76</span> <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">n_tokens</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">d_model</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">n_layers</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">layer</span><span class="p">:</span> <span class="n">TransformerLayer</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-14'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-14'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">83</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-15'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-15'>#</a>
|
||||
</div>
|
||||
<p>带<code class="highlight"><span></span><span class="n">n_layers</span></code>
|
||||
层的变压器</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">85</span> <span class="bp">self</span><span class="o">.</span><span class="n">transformer_layers</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">ModuleList</span><span class="p">([</span><span class="n">copy</span><span class="o">.</span><span class="n">deepcopy</span><span class="p">(</span><span class="n">layer</span><span class="p">)</span> <span class="k">for</span> <span class="n">_</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="n">n_layers</span><span class="p">)])</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-16'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-16'>#</a>
|
||||
</div>
|
||||
<p>令牌嵌入层</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">88</span> <span class="bp">self</span><span class="o">.</span><span class="n">emb</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Embedding</span><span class="p">(</span><span class="n">n_tokens</span><span class="p">,</span> <span class="n">d_model</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-17'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-17'>#</a>
|
||||
</div>
|
||||
<p>读出层</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">90</span> <span class="bp">self</span><span class="o">.</span><span class="n">readout</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Linear</span><span class="p">(</span><span class="n">d_model</span><span class="p">,</span> <span class="n">n_tokens</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-18'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-18'>#</a>
|
||||
</div>
|
||||
<p>掩码将在第一次调用时初始化</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">93</span> <span class="bp">self</span><span class="o">.</span><span class="n">mask</span> <span class="o">=</span> <span class="kc">None</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-19'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-19'>#</a>
|
||||
</div>
|
||||
<ul><li><code class="highlight"><span></span><span class="n">x</span></code>
|
||||
是形状的输入标记<code class="highlight"><span></span><span class="p">[</span><span class="n">seq_len</span><span class="p">,</span> <span class="n">batch_size</span><span class="p">]</span></code>
|
||||
</li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">95</span> <span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-20'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-20'>#</a>
|
||||
</div>
|
||||
<p>创建自动回归遮罩</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">100</span> <span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">mask</span> <span class="ow">is</span> <span class="kc">None</span> <span class="ow">or</span> <span class="bp">self</span><span class="o">.</span><span class="n">mask</span><span class="o">.</span><span class="n">size</span><span class="p">(</span><span class="mi">0</span><span class="p">)</span> <span class="o">!=</span> <span class="nb">len</span><span class="p">(</span><span class="n">x</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-21'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-21'>#</a>
|
||||
</div>
|
||||
<p>后续的掩码，将掩盖令牌以免看到未来的代币</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">102</span> <span class="bp">self</span><span class="o">.</span><span class="n">mask</span> <span class="o">=</span> <span class="n">subsequent_mask</span><span class="p">(</span><span class="nb">len</span><span class="p">(</span><span class="n">x</span><span class="p">))</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">x</span><span class="o">.</span><span class="n">device</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-22'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-22'>#</a>
|
||||
</div>
|
||||
<p>获取令牌嵌入</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">105</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">emb</span><span class="p">(</span><span class="n">x</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-23'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-23'>#</a>
|
||||
</div>
|
||||
<p>变压器编码</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">107</span> <span class="k">for</span> <span class="n">layer</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">transformer_layers</span><span class="p">:</span>
|
||||
<span class="lineno">108</span> <span class="n">x</span> <span class="o">=</span> <span class="n">layer</span><span class="p">(</span><span class="n">x</span><span class="o">=</span><span class="n">x</span><span class="p">,</span> <span class="n">mask</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">mask</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-24'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-24'>#</a>
|
||||
</div>
|
||||
<p>获取日志</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">110</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">readout</span><span class="p">(</span><span class="n">x</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-25'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-25'>#</a>
|
||||
</div>
|
||||
<p>返回结果</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">113</span> <span class="k">return</span> <span class="n">x</span><span class="p">,</span> <span class="kc">None</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-26'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-26'>#</a>
|
||||
</div>
|
||||
<h2>配置</h2>
|
||||
<p>这继承自 <a href="../../experiments/nlp_autoregression.html#NLPAutoRegressionConfigs"><code class="highlight"><span></span><span class="n">NLPAutoRegressionConfigs</span></code>
|
||||
</a></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">116</span><span class="k">class</span> <span class="nc">Configs</span><span class="p">(</span><span class="n">NLPAutoRegressionConfigs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-27'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-27'>#</a>
|
||||
</div>
|
||||
<p>型号</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">125</span> <span class="n">model</span><span class="p">:</span> <span class="n">AutoregressiveTransformer</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-28'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-28'>#</a>
|
||||
</div>
|
||||
<p>层数</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">128</span> <span class="n">n_layers</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">4</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-29'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-29'>#</a>
|
||||
</div>
|
||||
<p><span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.43056em;vertical-align:0em;"></span><span class="mord mathnormal" style="margin-right:0.0037em;">α</span></span></span></span></span>对<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.8888799999999999em;vertical-align:-0.19444em;"></span><span class="mord mathnormal" style="margin-right:0.05278em;">β</span></span></span></span></span>于 DeepNorm</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">131</span> <span class="n">deep_norm_alpha</span><span class="p">:</span> <span class="nb">float</span>
|
||||
<span class="lineno">132</span> <span class="n">deep_norm_beta</span><span class="p">:</span> <span class="nb">float</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-30'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-30'>#</a>
|
||||
</div>
|
||||
<p>关注的头部数量</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">135</span> <span class="n">n_heads</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">4</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-31'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-31'>#</a>
|
||||
</div>
|
||||
<p>嵌入大小</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">137</span> <span class="n">d_model</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">256</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-32'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-32'>#</a>
|
||||
</div>
|
||||
<p>每个注意头的大小</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">139</span> <span class="n">d_k</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">16</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-33'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-33'>#</a>
|
||||
</div>
|
||||
<p>前馈图层大小</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">141</span> <span class="n">d_ff</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">256</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-34'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-34'>#</a>
|
||||
</div>
|
||||
<p>自贸区</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">144</span> <span class="n">fta_lower_limit</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="o">-</span><span class="mf">1.</span>
|
||||
<span class="lineno">145</span> <span class="n">fta_upper_limit</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="o">+</span><span class="mf">1.</span>
|
||||
<span class="lineno">146</span> <span class="n">fta_delta</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">0.2</span>
|
||||
<span class="lineno">147</span> <span class="n">fta_eta</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">0.05</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-35'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-35'>#</a>
|
||||
</div>
|
||||
<h4>初始化模型</h4>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">150</span><span class="nd">@option</span><span class="p">(</span><span class="n">Configs</span><span class="o">.</span><span class="n">model</span><span class="p">)</span>
|
||||
<span class="lineno">151</span><span class="k">def</span> <span class="nf">_model</span><span class="p">(</span><span class="n">c</span><span class="p">:</span> <span class="n">Configs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-36'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-36'>#</a>
|
||||
</div>
|
||||
<p>创建 FTA 激活模块</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">157</span> <span class="n">fta</span> <span class="o">=</span> <span class="n">FTA</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">fta_lower_limit</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">fta_upper_limit</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">fta_delta</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">fta_eta</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-37'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-37'>#</a>
|
||||
</div>
|
||||
<p>创建变压器。我们重复使用<a href="../../transformers/models.html#TransformerLayer"><code class="highlight"><span></span><span class="n">TransformerLayer</span></code>
|
||||
</a>和<a href="../../transformers/mha.html"><code class="highlight"><span></span><span class="n">MultiHeadAttention</span></code>
|
||||
</a>实现。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">161</span> <span class="n">m</span> <span class="o">=</span> <span class="n">AutoregressiveTransformer</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">n_tokens</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">d_model</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">n_layers</span><span class="p">,</span>
|
||||
<span class="lineno">162</span> <span class="n">TransformerLayer</span><span class="p">(</span><span class="n">d_model</span><span class="o">=</span><span class="n">c</span><span class="o">.</span><span class="n">d_model</span><span class="p">,</span>
|
||||
<span class="lineno">163</span> <span class="n">feed_forward</span><span class="o">=</span><span class="n">FeedForwardFTA</span><span class="p">(</span><span class="n">d_model</span><span class="o">=</span><span class="n">c</span><span class="o">.</span><span class="n">d_model</span><span class="p">,</span>
|
||||
<span class="lineno">164</span> <span class="n">d_ff</span><span class="o">=</span><span class="n">c</span><span class="o">.</span><span class="n">d_ff</span><span class="p">,</span>
|
||||
<span class="lineno">165</span> <span class="n">activation</span><span class="o">=</span><span class="n">fta</span><span class="p">,</span>
|
||||
<span class="lineno">166</span> <span class="n">dropout</span><span class="o">=</span><span class="mf">0.1</span><span class="p">),</span>
|
||||
<span class="lineno">167</span> <span class="n">self_attn</span><span class="o">=</span><span class="n">MultiHeadAttention</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">n_heads</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">d_model</span><span class="p">,</span>
|
||||
<span class="lineno">168</span> <span class="n">dropout_prob</span><span class="o">=</span><span class="mf">0.0</span><span class="p">),</span>
|
||||
<span class="lineno">169</span> <span class="n">dropout_prob</span><span class="o">=</span><span class="mf">0.0</span><span class="p">))</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-38'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-38'>#</a>
|
||||
</div>
|
||||
<p>移到设备</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">172</span> <span class="k">return</span> <span class="n">m</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">device</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-39'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-39'>#</a>
|
||||
</div>
|
||||
<h4>创建并运行实验</h4>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">175</span><span class="k">def</span> <span class="nf">main</span><span class="p">():</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-40'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-40'>#</a>
|
||||
</div>
|
||||
<p>创建实验</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">180</span> <span class="n">experiment</span><span class="o">.</span><span class="n">create</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s2">"fta"</span><span class="p">,</span> <span class="n">writers</span><span class="o">=</span><span class="p">{</span><span class="s1">'screen'</span><span class="p">,</span> <span class="s1">'labml'</span><span class="p">})</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-41'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-41'>#</a>
|
||||
</div>
|
||||
<p>创建配置</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">182</span> <span class="n">conf</span> <span class="o">=</span> <span class="n">Configs</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-42'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-42'>#</a>
|
||||
</div>
|
||||
<p>覆盖配置</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">184</span> <span class="n">experiment</span><span class="o">.</span><span class="n">configs</span><span class="p">(</span><span class="n">conf</span><span class="p">,</span> <span class="p">{</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-43'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-43'>#</a>
|
||||
</div>
|
||||
<p>使用角色等级分词器</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">186</span> <span class="s1">'tokenizer'</span><span class="p">:</span> <span class="s1">'character'</span><span class="p">,</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-44'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-44'>#</a>
|
||||
</div>
|
||||
<p>提示分隔符为空</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">188</span> <span class="s1">'prompt_separator'</span><span class="p">:</span> <span class="s1">''</span><span class="p">,</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-45'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-45'>#</a>
|
||||
</div>
|
||||
<p>开始采样提示</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">190</span> <span class="s1">'prompt'</span><span class="p">:</span> <span class="s1">'It is '</span><span class="p">,</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-46'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-46'>#</a>
|
||||
</div>
|
||||
<p>使用小莎士比亚数据集</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">192</span> <span class="s1">'text'</span><span class="p">:</span> <span class="s1">'tiny_shakespeare'</span><span class="p">,</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-47'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-47'>#</a>
|
||||
</div>
|
||||
<p>使用上下文大小为<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord">256</span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">195</span> <span class="s1">'seq_len'</span><span class="p">:</span> <span class="mi">256</span><span class="p">,</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-48'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-48'>#</a>
|
||||
</div>
|
||||
<p>训练 32 个时代</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">197</span> <span class="s1">'epochs'</span><span class="p">:</span> <span class="mi">32</span><span class="p">,</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-49'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-49'>#</a>
|
||||
</div>
|
||||
<p>批量大小<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord">16</span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">199</span> <span class="s1">'batch_size'</span><span class="p">:</span> <span class="mi">16</span><span class="p">,</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-50'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-50'>#</a>
|
||||
</div>
|
||||
<p>在训练和验证之间切换每个纪元的<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord">10</span></span></span></span></span>次数</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">201</span> <span class="s1">'inner_iterations'</span><span class="p">:</span> <span class="mi">10</span><span class="p">,</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-51'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-51'>#</a>
|
||||
</div>
|
||||
<p>没有预热的 Adam 优化器</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">204</span> <span class="s1">'optimizer.optimizer'</span><span class="p">:</span> <span class="s1">'Adam'</span><span class="p">,</span>
|
||||
<span class="lineno">205</span> <span class="s1">'optimizer.learning_rate'</span><span class="p">:</span> <span class="mf">3e-4</span><span class="p">,</span>
|
||||
<span class="lineno">206</span> <span class="p">})</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-52'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-52'>#</a>
|
||||
</div>
|
||||
<p>设置用于保存和加载的模型</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">209</span> <span class="n">experiment</span><span class="o">.</span><span class="n">add_pytorch_models</span><span class="p">({</span><span class="s1">'model'</span><span class="p">:</span> <span class="n">conf</span><span class="o">.</span><span class="n">model</span><span class="p">})</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-53'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-53'>#</a>
|
||||
</div>
|
||||
<p>开始实验</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">212</span> <span class="k">with</span> <span class="n">experiment</span><span class="o">.</span><span class="n">start</span><span class="p">():</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-54'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-54'>#</a>
|
||||
</div>
|
||||
<p>跑步训练</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">214</span> <span class="n">conf</span><span class="o">.</span><span class="n">run</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-55'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-55'>#</a>
|
||||
</div>
|
||||
<p></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">218</span><span class="k">if</span> <span class="vm">__name__</span> <span class="o">==</span> <span class="s1">'__main__'</span><span class="p">:</span>
|
||||
<span class="lineno">219</span> <span class="n">main</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
@@ -0,0 +1,407 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content="PyTorch 实现和模糊平铺激活教程摘自《模糊平铺激活：在线学习稀疏表示的简单方法》一文。"/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="模糊平铺激活"/>
|
||||
<meta name="twitter:description" content="PyTorch 实现和模糊平铺激活教程摘自《模糊平铺激活：在线学习稀疏表示的简单方法》一文。"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/activations/fta/index.html"/>
|
||||
<meta property="og:title" content="模糊平铺激活"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="模糊平铺激活"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="模糊平铺激活"/>
|
||||
<meta property="og:description" content="PyTorch 实现和模糊平铺激活教程摘自《模糊平铺激活：在线学习稀疏表示的简单方法》一文。"/>
|
||||
|
||||
<title>模糊平铺激活</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/activations/fta/index.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="../index.html">activations</a>
|
||||
<a class="parent" href="index.html">fta</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/activations/fta/__init__.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1>模糊拼贴激活 (FTA)</h1>
|
||||
<p><a href="https://colab.research.google.com/github/labmlai/annotated_deep_learning_paper_implementations/blob/master/labml_nn/activations/fta/experiment.ipynb"><img alt="Open In Colab" src="https://colab.research.google.com/assets/colab-badge.svg"></a></p>
|
||||
<p>这是 <a href="https://pytorch.org">PyTorch</a> 的 <a href="https://arxiv.org/abs/1911.08068">Fuzzy Tiling Activations 实现/教程：一种在线学习稀疏表示的简单方法</a>。</p>
|
||||
<p>模糊平铺激活是一种基于分箱的稀疏激活形式。</p>
|
||||
分@@ <p>箱是根据间隔将标量值分类到数据桶中。分箱的一个问题是，它为大多数值提供零梯度（区间边界除外）。另一个原因是，如果分箱间隔很大，则分箱会失去精度。</p>
|
||||
<p>自由贸易协定克服了这些缺点。FTA 没有像 Tiling Activations 那样使用硬边界，而是使用箱子之间的软边界。这为所有或很宽范围的值提供了非零梯度。而且也不会损失精度，因为它是在部分值中捕获的。</p>
|
||||
<h4>平铺激活</h4>
|
||||
<p><span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.44444em;vertical-align:0em;"></span><span class="mord coloredeq eqj" style=""><span class="mord mathbf" style="">c</span></span></span></span></span></span>是平铺向量，</p>
|
||||
<p><span ><span class="katex-display"><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:1em;vertical-align:-0.25em;"></span><span class="mord coloredeq eqc" style=""><span class="mord" style=""><span class="mord mathbf coloredeq eqj" style="">c</span></span><span class="mspace" style="margin-right:0.2777777777777778em;"></span><span class="mrel" style="">=</span><span class="mspace" style="margin-right:0.2777777777777778em;"></span><span class="mopen" style="">(</span><span class="mord" style=""><span class="mord mathnormal coloredeq eqq" style="margin-right:0.01968em">l</span></span><span class="mpunct" style="">,</span><span class="mspace" style="margin-right:0.16666666666666666em;"></span><span class="mord" style=""><span class="mord mathnormal coloredeq eqq" style="margin-right:0.01968em">l</span></span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">+</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord mathnormal" style="margin-right:0.03785em">δ</span><span class="mpunct" style="">,</span><span class="mspace" style="margin-right:0.16666666666666666em;"></span><span class="mord" style=""><span class="mord mathnormal coloredeq eqq" style="margin-right:0.01968em">l</span></span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">+</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord" style="">2</span><span class="mord mathnormal" style="margin-right:0.03785em">δ</span><span class="mpunct" style="">,</span><span class="mspace" style="margin-right:0.16666666666666666em;"></span><span class="minner" style="">…</span><span class="mspace" style="margin-right:0.16666666666666666em;"></span><span class="mpunct" style="">,</span><span class="mspace" style="margin-right:0.16666666666666666em;"></span><span class="mord" style=""><span class="mord mathnormal coloredeq eqr" style="">u</span></span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">−</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord" style="">2</span><span class="mord mathnormal" style="margin-right:0.03785em">δ</span><span class="mpunct" style="">,</span><span class="mspace" style="margin-right:0.16666666666666666em;"></span><span class="mord" style=""><span class="mord mathnormal coloredeq eqr" style="">u</span></span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">−</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord" style=""><span class="mord mathnormal coloredeq eql" style="margin-right:0.03785em">δ</span></span><span class="mclose" style="">)</span></span></span></span></span></span></span></p>
|
||||
<p>其中，<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:1em;vertical-align:-0.25em;"></span><span class="mopen">[</span><span class="mord coloredeq eqq" style=""><span class="mord mathnormal" style="margin-right:0.01968em">l</span></span><span class="mpunct">,</span><span class="mspace" style="margin-right:0.16666666666666666em;"></span><span class="mord coloredeq eqr" style=""><span class="mord mathnormal" style="">u</span></span><span class="mclose">]</span></span></span></span></span>是输入范围，<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.69444em;vertical-align:0em;"></span><span class="mord coloredeq eql" style=""><span class="mord mathnormal" style="margin-right:0.03785em">δ</span></span></span></span></span></span>是图格大小，<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.77777em;vertical-align:-0.08333em;"></span><span class="mord coloredeq eqm" style=""><span class="mord" style=""><span class="mord mathnormal coloredeq eqr" style="">u</span></span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">−</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord" style=""><span class="mord mathnormal coloredeq eqq" style="margin-right:0.01968em">l</span></span></span></span></span></span></span>可被整除<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.69444em;vertical-align:0em;"></span><span class="mord coloredeq eql" style=""><span class="mord mathnormal" style="margin-right:0.03785em">δ</span></span></span></span></span></span>。</p>
|
||||
<p>平铺激活是，</p>
|
||||
<p><span ><span class="katex-display"><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:1em;vertical-align:-0.25em;"></span><span class="mord mathnormal">ϕ</span><span class="mopen">(</span><span class="mord coloredeq eqs" style=""><span class="mord mathnormal" style="margin-right:0.04398em">z</span></span><span class="mclose">)</span><span class="mspace" style="margin-right:0.2777777777777778em;"></span><span class="mrel">=</span><span class="mspace" style="margin-right:0.2777777777777778em;"></span></span><span class="base"><span class="strut" style="height:0.72777em;vertical-align:-0.08333em;"></span><span class="mord coloredeq eqp" style=""><span class="mord" style="">1</span></span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin">−</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span></span><span class="base"><span class="strut" style="height:1.20001em;vertical-align:-0.35001em;"></span><span class="mord"><span class="mord mathnormal" style="margin-right:0.07847em;">I</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.25833100000000003em;"><span style="top:-2.5500000000000003em;margin-left:-0.07847em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight"><span class="mbin mtight">+</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.208331em;"><span></span></span></span></span></span></span><span class="mord"><span class="delimsizing size1">(</span></span><span class="mspace" style="margin-right:0.16666666666666666em;"></span><span class="mop">max</span><span class="mopen">(</span><span class="mord coloredeq eqj" style=""><span class="mord mathbf" style="">c</span></span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin">−</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span></span><span class="base"><span class="strut" style="height:1em;vertical-align:-0.25em;"></span><span class="mord coloredeq eqs" style=""><span class="mord mathnormal" style="margin-right:0.04398em">z</span></span><span class="mpunct">,</span><span class="mspace" style="margin-right:0.16666666666666666em;"></span><span class="mord coloredeq eqo" style=""><span class="mord" style="">0</span></span><span class="mclose">)</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin">+</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span></span><span class="base"><span class="strut" style="height:1em;vertical-align:-0.25em;"></span><span class="mop">max</span><span class="mopen">(</span><span class="mord coloredeq eqs" style=""><span class="mord mathnormal" style="margin-right:0.04398em">z</span></span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin">−</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span></span><span class="base"><span class="strut" style="height:0.77777em;vertical-align:-0.08333em;"></span><span class="mord coloredeq eql" style=""><span class="mord mathnormal" style="margin-right:0.03785em">δ</span></span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin">−</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span></span><span class="base"><span class="strut" style="height:1.20001em;vertical-align:-0.35001em;"></span><span class="mord coloredeq eqj" style=""><span class="mord mathbf" style="">c</span></span><span class="mclose">)</span><span class="mord"><span class="delimsizing size1">)</span></span></span></span></span></span></span></p>
|
||||
<p>其中<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:1em;vertical-align:-0.25em;"></span><span class="mord"><span class="mord mathnormal" style="margin-right:0.07847em;">I</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.25833100000000003em;"><span style="top:-2.5500000000000003em;margin-left:-0.07847em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight"><span class="mbin mtight">+</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.208331em;"><span></span></span></span></span></span></span><span class="mopen">(</span><span class="mord">⋅</span><span class="mclose">)</span></span></span></span></span>，是指示函数，该函数给出输入<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqp" style=""><span class="mord" style="">1</span></span></span></span></span></span>是否为正，<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqo" style=""><span class="mord" style="">0</span></span></span></span></span></span>否则。</p>
|
||||
<p>请注意，平铺激活的梯度为零，因为它有硬边界。</p>
|
||||
<h4>模糊拼贴激活</h4>
|
||||
<p>模糊指标函数，</p>
|
||||
<p><span ><span class="katex-display"><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:1.036108em;vertical-align:-0.286108em;"></span><span class="mord coloredeq eqd" style=""><span class="mord" style=""><span class="mord mathnormal" style="margin-right:0.07847em">I</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.25833100000000003em;"><span style="top:-2.5500000000000003em;margin-left:-0.07847em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style=""><span class="mord mathnormal mtight" style="margin-right:0.03588em">η</span><span class="mpunct mtight" style="">,</span><span class="mord mtight" style="">+</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.286108em;"><span></span></span></span></span></span></span><span class="mopen" style="">(</span><span class="mord mathnormal" style="">x</span><span class="mclose" style="">)</span><span class="mspace" style="margin-right:0.2777777777777778em;"></span><span class="mrel" style="">=</span><span class="mspace" style="margin-right:0.2777777777777778em;"></span><span class="mord" style=""><span class="mord mathnormal" style="margin-right:0.07847em">I</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.25833100000000003em;"><span style="top:-2.5500000000000003em;margin-left:-0.07847em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mbin mtight" style="">+</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.208331em;"><span></span></span></span></span></span></span><span class="mopen" style="">(</span><span class="mord" style=""><span class="mord mathnormal coloredeq eqn" style="margin-right:0.03588em">η</span></span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">−</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord mathnormal" style="">x</span><span class="mclose" style="">)</span><span class="mord mathnormal" style="">x</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">+</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord" style=""><span class="mord mathnormal" style="margin-right:0.07847em">I</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.25833100000000003em;"><span style="top:-2.5500000000000003em;margin-left:-0.07847em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mbin mtight" style="">+</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.208331em;"><span></span></span></span></span></span></span><span class="mopen" style="">(</span><span class="mord mathnormal" style="">x</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">−</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord" style=""><span class="mord mathnormal coloredeq eqn" style="margin-right:0.03588em">η</span></span><span class="mclose" style="">)</span></span></span></span></span></span></span></p>
|
||||
<p>它从<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqo" style=""><span class="mord" style="">0</span></span></span></span></span></span>到<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqp" style=""><span class="mord" style="">1</span></span></span></span></span></span>何时线性增加<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.78041em;vertical-align:-0.13597em;"></span><span class="mord coloredeq eqo" style=""><span class="mord" style="">0</span></span><span class="mspace" style="margin-right:0.2777777777777778em;"></span><span class="mrel">≤</span><span class="mspace" style="margin-right:0.2777777777777778em;"></span></span><span class="base"><span class="strut" style="height:0.5782em;vertical-align:-0.0391em;"></span><span class="mord mathnormal">x</span><span class="mspace" style="margin-right:0.2777777777777778em;"></span><span class="mrel"><</span><span class="mspace" style="margin-right:0.2777777777777778em;"></span></span><span class="base"><span class="strut" style="height:0.625em;vertical-align:-0.19444em;"></span><span class="mord coloredeq eqn" style=""><span class="mord mathnormal" style="margin-right:0.03588em">η</span></span></span></span></span></span>，等<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqp" style=""><span class="mord" style="">1</span></span></span></span></span></span>于 for<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.8304100000000001em;vertical-align:-0.19444em;"></span><span class="mord coloredeq eqn" style=""><span class="mord mathnormal" style="margin-right:0.03588em">η</span></span><span class="mspace" style="margin-right:0.2777777777777778em;"></span><span class="mrel">≤</span><span class="mspace" style="margin-right:0.2777777777777778em;"></span></span><span class="base"><span class="strut" style="height:0.43056em;vertical-align:0em;"></span><span class="mord mathnormal">x</span></span></span></span></span>。<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.625em;vertical-align:-0.19444em;"></span><span class="mord coloredeq eqn" style=""><span class="mord mathnormal" style="margin-right:0.03588em">η</span></span></span></span></span></span>是一个超参数。</p>
|
||||
<p>FTA 使用它在箱子之间创建软边界。</p>
|
||||
<p><span ><span class="katex-display"><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:1.20001em;vertical-align:-0.35001em;"></span><span class="mord coloredeq eqa" style=""><span class="mord" style=""><span class="mord coloredeq eqg" style=""><span class="mord mathnormal" style="">ϕ</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.15139200000000003em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style=""><span class="mord mathnormal mtight coloredeq eqn" style="margin-right:0.03588em">η</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.286108em;"><span></span></span></span></span></span></span><span class="mopen coloredeq eqg" style="">(</span><span class="mord coloredeq eqg" style=""><span class="mord mathnormal coloredeq eqs" style="margin-right:0.04398em">z</span></span><span class="mclose coloredeq eqg" style="">)</span></span><span class="mspace" style="margin-right:0.2777777777777778em;"></span><span class="mrel" style="">=</span><span class="mspace" style="margin-right:0.2777777777777778em;"></span><span class="mord" style=""><span class="mord coloredeq eqp" style="">1</span></span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">−</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord" style=""><span class="mord mathnormal" style="margin-right:0.07847em">I</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.25833100000000003em;"><span style="top:-2.5500000000000003em;margin-left:-0.07847em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style=""><span class="mord mathnormal mtight" style="margin-right:0.03588em">η</span><span class="mpunct mtight" style="">,</span><span class="mord mtight" style="">+</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.286108em;"><span></span></span></span></span></span></span><span class="mord" style=""><span class="delimsizing size1" style=""><span style="">(</span></span></span><span class="mspace" style="margin-right:0.16666666666666666em;"></span><span class="mop" style=""><span style="">m</span><span style="">a</span><span style="">x</span></span><span class="mopen" style="">(</span><span class="mord" style=""><span class="mord mathbf coloredeq eqj" style="">c</span></span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">−</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord" style=""><span class="mord mathnormal coloredeq eqs" style="margin-right:0.04398em">z</span></span><span class="mpunct" style="">,</span><span class="mspace" style="margin-right:0.16666666666666666em;"></span><span class="mord" style=""><span class="mord coloredeq eqo" style="">0</span></span><span class="mclose" style="">)</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">+</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mop" style=""><span style="">m</span><span style="">a</span><span style="">x</span></span><span class="mopen" style="">(</span><span class="mord" style=""><span class="mord mathnormal coloredeq eqs" style="margin-right:0.04398em">z</span></span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">−</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord" style=""><span class="mord mathnormal coloredeq eql" style="margin-right:0.03785em">δ</span></span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">−</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord" style=""><span class="mord mathbf coloredeq eqj" style="">c</span></span><span class="mpunct" style="">,</span><span class="mspace" style="margin-right:0.16666666666666666em;"></span><span class="mord" style=""><span class="mord coloredeq eqo" style="">0</span></span><span class="mclose" style="">)</span><span class="mord" style=""><span class="delimsizing size1" style=""><span style="">)</span></span></span></span></span></span></span></span></span></p>
|
||||
<p><a href="experiment.html">这是一个在变压器中使用 FTA 的简单实验</a>。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">61</span><span></span><span class="kn">import</span> <span class="nn">torch</span>
|
||||
<span class="lineno">62</span><span class="kn">from</span> <span class="nn">torch</span> <span class="kn">import</span> <span class="n">nn</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-1'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-1'>#</a>
|
||||
</div>
|
||||
<h3>模糊平铺激活 (FTA)</h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">65</span><span class="k">class</span> <span class="nc">FTA</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-2'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-2'>#</a>
|
||||
</div>
|
||||
<ul><li><code class="highlight"><span></span><span class="n">lower_limit</span></code>
|
||||
是下限<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.69444em;vertical-align:0em;"></span><span class="mord coloredeq eqq" style=""><span class="mord mathnormal" style="margin-right:0.01968em">l</span></span></span></span></span></span></li>
|
||||
<li><code class="highlight"><span></span><span class="n">upper_limit</span></code>
|
||||
是上限<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.43056em;vertical-align:0em;"></span><span class="mord coloredeq eqr" style=""><span class="mord mathnormal" style="">u</span></span></span></span></span></span></li>
|
||||
<li><code class="highlight"><span></span><span class="n">delta</span></code>
|
||||
是垃圾桶的大小<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.69444em;vertical-align:0em;"></span><span class="mord coloredeq eql" style=""><span class="mord mathnormal" style="margin-right:0.03785em">δ</span></span></span></span></span></span></li>
|
||||
<li><code class="highlight"><span></span><span class="n">eta</span></code>
|
||||
<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.625em;vertical-align:-0.19444em;"></span><span class="mord coloredeq eqn" style=""><span class="mord mathnormal" style="margin-right:0.03588em">η</span></span></span></span></span></span>是确定边界柔和度的参数。</li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">70</span> <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">lower_limit</span><span class="p">:</span> <span class="nb">float</span><span class="p">,</span> <span class="n">upper_limit</span><span class="p">:</span> <span class="nb">float</span><span class="p">,</span> <span class="n">delta</span><span class="p">:</span> <span class="nb">float</span><span class="p">,</span> <span class="n">eta</span><span class="p">:</span> <span class="nb">float</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-3'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-3'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">77</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-4'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-4'>#</a>
|
||||
</div>
|
||||
<p>初始化平铺矢量<span ><span class="katex-display"><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:1em;vertical-align:-0.25em;"></span><span class="mord coloredeq eqc" style=""><span class="mord" style=""><span class="mord mathbf coloredeq eqj" style="">c</span></span><span class="mspace" style="margin-right:0.2777777777777778em;"></span><span class="mrel" style="">=</span><span class="mspace" style="margin-right:0.2777777777777778em;"></span><span class="mopen" style="">(</span><span class="mord" style=""><span class="mord mathnormal coloredeq eqq" style="margin-right:0.01968em">l</span></span><span class="mpunct" style="">,</span><span class="mspace" style="margin-right:0.16666666666666666em;"></span><span class="mord" style=""><span class="mord mathnormal coloredeq eqq" style="margin-right:0.01968em">l</span></span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">+</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord mathnormal" style="margin-right:0.03785em">δ</span><span class="mpunct" style="">,</span><span class="mspace" style="margin-right:0.16666666666666666em;"></span><span class="mord" style=""><span class="mord mathnormal coloredeq eqq" style="margin-right:0.01968em">l</span></span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">+</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord" style="">2</span><span class="mord mathnormal" style="margin-right:0.03785em">δ</span><span class="mpunct" style="">,</span><span class="mspace" style="margin-right:0.16666666666666666em;"></span><span class="minner" style="">…</span><span class="mspace" style="margin-right:0.16666666666666666em;"></span><span class="mpunct" style="">,</span><span class="mspace" style="margin-right:0.16666666666666666em;"></span><span class="mord" style=""><span class="mord mathnormal coloredeq eqr" style="">u</span></span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">−</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord" style="">2</span><span class="mord mathnormal" style="margin-right:0.03785em">δ</span><span class="mpunct" style="">,</span><span class="mspace" style="margin-right:0.16666666666666666em;"></span><span class="mord" style=""><span class="mord mathnormal coloredeq eqr" style="">u</span></span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">−</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord" style=""><span class="mord mathnormal coloredeq eql" style="margin-right:0.03785em">δ</span></span><span class="mclose" style="">)</span></span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">80</span> <span class="bp">self</span><span class="o">.</span><span class="n">c</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Parameter</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">arange</span><span class="p">(</span><span class="n">lower_limit</span><span class="p">,</span> <span class="n">upper_limit</span><span class="p">,</span> <span class="n">delta</span><span class="p">),</span> <span class="n">requires_grad</span><span class="o">=</span><span class="kc">False</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-5'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-5'>#</a>
|
||||
</div>
|
||||
<p>输入向量展开的系数等于条柱的数量<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:1.2251079999999999em;vertical-align:-0.345em;"></span><span class="mord coloredeq eqe" style=""><span class="mord" style=""><span class="mopen nulldelimiter"></span><span class="mfrac"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.8801079999999999em;"><span style="top:-2.6550000000000002em;"><span class="pstrut" style="height:3em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style=""><span class="mord mtight" style=""><span class="mord mathnormal mtight coloredeq eql" style="margin-right:0.03785em">δ</span></span></span></span></span><span style="top:-3.23em;"><span class="pstrut" style="height:3em;"></span><span class="frac-line" style="border-bottom-width:0.04em"></span></span><span style="top:-3.394em;"><span class="pstrut" style="height:3em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style=""><span class="mord mtight" style=""><span class="mord mtight coloredeq eqm" style=""><span class="mord mathnormal mtight coloredeq eqr" style="">u</span></span><span class="mbin mtight coloredeq eqm" style="">−</span><span class="mord mtight coloredeq eqm" style=""><span class="mord mathnormal mtight coloredeq eqq" style="margin-right:0.01968em">l</span></span></span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.345em;"><span></span></span></span></span></span><span class="mclose nulldelimiter"></span></span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">82</span> <span class="bp">self</span><span class="o">.</span><span class="n">expansion_factor</span> <span class="o">=</span> <span class="nb">len</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">c</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-6'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-6'>#</a>
|
||||
</div>
|
||||
<p><span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.69444em;vertical-align:0em;"></span><span class="mord coloredeq eql" style=""><span class="mord mathnormal" style="margin-right:0.03785em">δ</span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">84</span> <span class="bp">self</span><span class="o">.</span><span class="n">delta</span> <span class="o">=</span> <span class="n">delta</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-7'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-7'>#</a>
|
||||
</div>
|
||||
<p><span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.625em;vertical-align:-0.19444em;"></span><span class="mord coloredeq eqn" style=""><span class="mord mathnormal" style="margin-right:0.03588em">η</span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">86</span> <span class="bp">self</span><span class="o">.</span><span class="n">eta</span> <span class="o">=</span> <span class="n">eta</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-8'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-8'>#</a>
|
||||
</div>
|
||||
<h4>模糊指标函数</h4>
|
||||
<p><span ><span class="katex-display"><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:1.036108em;vertical-align:-0.286108em;"></span><span class="mord coloredeq eqd" style=""><span class="mord" style=""><span class="mord mathnormal" style="margin-right:0.07847em">I</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.25833100000000003em;"><span style="top:-2.5500000000000003em;margin-left:-0.07847em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style=""><span class="mord mathnormal mtight" style="margin-right:0.03588em">η</span><span class="mpunct mtight" style="">,</span><span class="mord mtight" style="">+</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.286108em;"><span></span></span></span></span></span></span><span class="mopen" style="">(</span><span class="mord mathnormal" style="">x</span><span class="mclose" style="">)</span><span class="mspace" style="margin-right:0.2777777777777778em;"></span><span class="mrel" style="">=</span><span class="mspace" style="margin-right:0.2777777777777778em;"></span><span class="mord" style=""><span class="mord mathnormal" style="margin-right:0.07847em">I</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.25833100000000003em;"><span style="top:-2.5500000000000003em;margin-left:-0.07847em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mbin mtight" style="">+</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.208331em;"><span></span></span></span></span></span></span><span class="mopen" style="">(</span><span class="mord" style=""><span class="mord mathnormal coloredeq eqn" style="margin-right:0.03588em">η</span></span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">−</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord mathnormal" style="">x</span><span class="mclose" style="">)</span><span class="mord mathnormal" style="">x</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">+</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord" style=""><span class="mord mathnormal" style="margin-right:0.07847em">I</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.25833100000000003em;"><span style="top:-2.5500000000000003em;margin-left:-0.07847em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mbin mtight" style="">+</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.208331em;"><span></span></span></span></span></span></span><span class="mopen" style="">(</span><span class="mord mathnormal" style="">x</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">−</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord" style=""><span class="mord mathnormal coloredeq eqn" style="margin-right:0.03588em">η</span></span><span class="mclose" style="">)</span></span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">88</span> <span class="k">def</span> <span class="nf">fuzzy_i_plus</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-9'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-9'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">94</span> <span class="k">return</span> <span class="p">(</span><span class="n">x</span> <span class="o"><=</span> <span class="bp">self</span><span class="o">.</span><span class="n">eta</span><span class="p">)</span> <span class="o">*</span> <span class="n">x</span> <span class="o">+</span> <span class="p">(</span><span class="n">x</span> <span class="o">></span> <span class="bp">self</span><span class="o">.</span><span class="n">eta</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-10'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-10'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">96</span> <span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">z</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-11'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-11'>#</a>
|
||||
</div>
|
||||
<p>再添加一个尺寸尺寸<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqp" style=""><span class="mord" style="">1</span></span></span></span></span></span>。我们会将其扩展为垃圾箱。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">99</span> <span class="n">z</span> <span class="o">=</span> <span class="n">z</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="o">*</span><span class="n">z</span><span class="o">.</span><span class="n">shape</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-12'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-12'>#</a>
|
||||
</div>
|
||||
<p><span ><span class="katex-display"><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:1.20001em;vertical-align:-0.35001em;"></span><span class="mord coloredeq eqa" style=""><span class="mord" style=""><span class="mord coloredeq eqg" style=""><span class="mord mathnormal" style="">ϕ</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.15139200000000003em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style=""><span class="mord mathnormal mtight coloredeq eqn" style="margin-right:0.03588em">η</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.286108em;"><span></span></span></span></span></span></span><span class="mopen coloredeq eqg" style="">(</span><span class="mord coloredeq eqg" style=""><span class="mord mathnormal coloredeq eqs" style="margin-right:0.04398em">z</span></span><span class="mclose coloredeq eqg" style="">)</span></span><span class="mspace" style="margin-right:0.2777777777777778em;"></span><span class="mrel" style="">=</span><span class="mspace" style="margin-right:0.2777777777777778em;"></span><span class="mord" style=""><span class="mord coloredeq eqp" style="">1</span></span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">−</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord" style=""><span class="mord mathnormal" style="margin-right:0.07847em">I</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.25833100000000003em;"><span style="top:-2.5500000000000003em;margin-left:-0.07847em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style=""><span class="mord mathnormal mtight" style="margin-right:0.03588em">η</span><span class="mpunct mtight" style="">,</span><span class="mord mtight" style="">+</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.286108em;"><span></span></span></span></span></span></span><span class="mord" style=""><span class="delimsizing size1" style=""><span style="">(</span></span></span><span class="mspace" style="margin-right:0.16666666666666666em;"></span><span class="mop" style=""><span style="">m</span><span style="">a</span><span style="">x</span></span><span class="mopen" style="">(</span><span class="mord" style=""><span class="mord mathbf coloredeq eqj" style="">c</span></span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">−</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord" style=""><span class="mord mathnormal coloredeq eqs" style="margin-right:0.04398em">z</span></span><span class="mpunct" style="">,</span><span class="mspace" style="margin-right:0.16666666666666666em;"></span><span class="mord" style=""><span class="mord coloredeq eqo" style="">0</span></span><span class="mclose" style="">)</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">+</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mop" style=""><span style="">m</span><span style="">a</span><span style="">x</span></span><span class="mopen" style="">(</span><span class="mord" style=""><span class="mord mathnormal coloredeq eqs" style="margin-right:0.04398em">z</span></span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">−</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord" style=""><span class="mord mathnormal coloredeq eql" style="margin-right:0.03785em">δ</span></span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">−</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord" style=""><span class="mord mathbf coloredeq eqj" style="">c</span></span><span class="mpunct" style="">,</span><span class="mspace" style="margin-right:0.16666666666666666em;"></span><span class="mord" style=""><span class="mord coloredeq eqo" style="">0</span></span><span class="mclose" style="">)</span><span class="mord" style=""><span class="delimsizing size1" style=""><span style="">)</span></span></span></span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">102</span> <span class="n">z</span> <span class="o">=</span> <span class="mf">1.</span> <span class="o">-</span> <span class="bp">self</span><span class="o">.</span><span class="n">fuzzy_i_plus</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">clip</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">c</span> <span class="o">-</span> <span class="n">z</span><span class="p">,</span> <span class="nb">min</span><span class="o">=</span><span class="mf">0.</span><span class="p">)</span> <span class="o">+</span> <span class="n">torch</span><span class="o">.</span><span class="n">clip</span><span class="p">(</span><span class="n">z</span> <span class="o">-</span> <span class="bp">self</span><span class="o">.</span><span class="n">delta</span> <span class="o">-</span> <span class="bp">self</span><span class="o">.</span><span class="n">c</span><span class="p">,</span> <span class="nb">min</span><span class="o">=</span><span class="mf">0.</span><span class="p">))</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-13'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-13'>#</a>
|
||||
</div>
|
||||
<p>重塑为原始尺寸数。最后一个维度大小将按条柱的数量进行扩展<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:1.2251079999999999em;vertical-align:-0.345em;"></span><span class="mord coloredeq eqe" style=""><span class="mord" style=""><span class="mopen nulldelimiter"></span><span class="mfrac"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.8801079999999999em;"><span style="top:-2.6550000000000002em;"><span class="pstrut" style="height:3em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style=""><span class="mord mtight" style=""><span class="mord mathnormal mtight coloredeq eql" style="margin-right:0.03785em">δ</span></span></span></span></span><span style="top:-3.23em;"><span class="pstrut" style="height:3em;"></span><span class="frac-line" style="border-bottom-width:0.04em"></span></span><span style="top:-3.394em;"><span class="pstrut" style="height:3em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style=""><span class="mord mtight" style=""><span class="mord mtight coloredeq eqm" style=""><span class="mord mathnormal mtight coloredeq eqr" style="">u</span></span><span class="mbin mtight coloredeq eqm" style="">−</span><span class="mord mtight coloredeq eqm" style=""><span class="mord mathnormal mtight coloredeq eqq" style="margin-right:0.01968em">l</span></span></span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.345em;"><span></span></span></span></span></span><span class="mclose nulldelimiter"></span></span></span></span></span></span></span>。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">106</span> <span class="k">return</span> <span class="n">z</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="o">*</span><span class="n">z</span><span class="o">.</span><span class="n">shape</span><span class="p">[:</span><span class="o">-</span><span class="mi">2</span><span class="p">],</span> <span class="o">-</span><span class="mi">1</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-14'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-14'>#</a>
|
||||
</div>
|
||||
<h4>用于测试 FTA 模块的代码</h4>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">109</span><span class="k">def</span> <span class="nf">_test</span><span class="p">():</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-15'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-15'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">113</span> <span class="kn">from</span> <span class="nn">labml.logger</span> <span class="kn">import</span> <span class="n">inspect</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-16'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-16'>#</a>
|
||||
</div>
|
||||
<p>初始化</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">116</span> <span class="n">a</span> <span class="o">=</span> <span class="n">FTA</span><span class="p">(</span><span class="o">-</span><span class="mi">10</span><span class="p">,</span> <span class="mi">10</span><span class="p">,</span> <span class="mf">2.</span><span class="p">,</span> <span class="mf">0.5</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-17'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-17'>#</a>
|
||||
</div>
|
||||
<p>打印<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.44444em;vertical-align:0em;"></span><span class="mord coloredeq eqj" style=""><span class="mord mathbf" style="">c</span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">118</span> <span class="n">inspect</span><span class="p">(</span><span class="n">a</span><span class="o">.</span><span class="n">c</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-18'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-18'>#</a>
|
||||
</div>
|
||||
<p>打印纸箱数量<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:1.2251079999999999em;vertical-align:-0.345em;"></span><span class="mord coloredeq eqe" style=""><span class="mord" style=""><span class="mopen nulldelimiter"></span><span class="mfrac"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.8801079999999999em;"><span style="top:-2.6550000000000002em;"><span class="pstrut" style="height:3em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style=""><span class="mord mtight" style=""><span class="mord mathnormal mtight coloredeq eql" style="margin-right:0.03785em">δ</span></span></span></span></span><span style="top:-3.23em;"><span class="pstrut" style="height:3em;"></span><span class="frac-line" style="border-bottom-width:0.04em"></span></span><span style="top:-3.394em;"><span class="pstrut" style="height:3em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style=""><span class="mord mtight" style=""><span class="mord mtight coloredeq eqm" style=""><span class="mord mathnormal mtight coloredeq eqr" style="">u</span></span><span class="mbin mtight coloredeq eqm" style="">−</span><span class="mord mtight coloredeq eqm" style=""><span class="mord mathnormal mtight coloredeq eqq" style="margin-right:0.01968em">l</span></span></span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.345em;"><span></span></span></span></span></span><span class="mclose nulldelimiter"></span></span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">120</span> <span class="n">inspect</span><span class="p">(</span><span class="n">a</span><span class="o">.</span><span class="n">expansion_factor</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-19'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-19'>#</a>
|
||||
</div>
|
||||
<p>输入<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.43056em;vertical-align:0em;"></span><span class="mord coloredeq eqs" style=""><span class="mord mathnormal" style="margin-right:0.04398em">z</span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">123</span> <span class="n">z</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">tensor</span><span class="p">([</span><span class="mf">1.1</span><span class="p">,</span> <span class="mf">2.2</span><span class="p">,</span> <span class="mf">3.3</span><span class="p">,</span> <span class="mf">4.4</span><span class="p">,</span> <span class="mf">5.5</span><span class="p">,</span> <span class="mf">6.6</span><span class="p">,</span> <span class="mf">7.7</span><span class="p">,</span> <span class="mf">8.8</span><span class="p">,</span> <span class="mf">9.</span><span class="p">,</span> <span class="mf">10.</span><span class="p">,</span> <span class="mf">11.</span><span class="p">])</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-20'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-20'>#</a>
|
||||
</div>
|
||||
<p>打印<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.43056em;vertical-align:0em;"></span><span class="mord coloredeq eqs" style=""><span class="mord mathnormal" style="margin-right:0.04398em">z</span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">125</span> <span class="n">inspect</span><span class="p">(</span><span class="n">z</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-21'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-21'>#</a>
|
||||
</div>
|
||||
<p>打印<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:1.036108em;vertical-align:-0.286108em;"></span><span class="mord coloredeq eqg" style=""><span class="mord" style=""><span class="mord mathnormal" style="">ϕ</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.15139200000000003em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style=""><span class="mord mathnormal mtight coloredeq eqn" style="margin-right:0.03588em">η</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.286108em;"><span></span></span></span></span></span></span><span class="mopen" style="">(</span><span class="mord" style=""><span class="mord mathnormal coloredeq eqs" style="margin-right:0.04398em">z</span></span><span class="mclose" style="">)</span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">127</span> <span class="n">inspect</span><span class="p">(</span><span class="n">a</span><span class="p">(</span><span class="n">z</span><span class="p">))</span>
|
||||
<span class="lineno">128</span>
|
||||
<span class="lineno">129</span>
|
||||
<span class="lineno">130</span><span class="k">if</span> <span class="vm">__name__</span> <span class="o">==</span> <span class="s1">'__main__'</span><span class="p">:</span>
|
||||
<span class="lineno">131</span> <span class="n">_test</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
@@ -0,0 +1,127 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content="一组与神经网络激活相关的 PyTorch 实现/教程"/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="神经网络激活函数"/>
|
||||
<meta name="twitter:description" content="一组与神经网络激活相关的 PyTorch 实现/教程"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/activations/index.html"/>
|
||||
<meta property="og:title" content="神经网络激活函数"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="神经网络激活函数"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="神经网络激活函数"/>
|
||||
<meta property="og:description" content="一组与神经网络激活相关的 PyTorch 实现/教程"/>
|
||||
|
||||
<title>神经网络激活函数</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/activations/index.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="index.html">activations</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/activations/__init__.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1>神经网络激活</h1>
|
||||
<ul><li><a href="fta/index.html">模糊平铺激活</a></li>
|
||||
<li>🚧 <a href="swish/index.html">Swish</a></li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">14</span><span></span><span class="kn">from</span> <span class="nn">.swish</span> <span class="kn">import</span> <span class="n">Swish</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
@@ -0,0 +1,163 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content=""/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="swish.py"/>
|
||||
<meta name="twitter:description" content=""/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/activations/swish.html"/>
|
||||
<meta property="og:title" content="swish.py"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="swish.py"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="swish.py"/>
|
||||
<meta property="og:description" content=""/>
|
||||
|
||||
<title>swish.py</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/activations/swish.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="index.html">activations</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/activations/swish.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">1</span><span></span><span class="kn">import</span> <span class="nn">torch</span>
|
||||
<span class="lineno">2</span><span class="kn">from</span> <span class="nn">torch</span> <span class="kn">import</span> <span class="n">nn</span>
|
||||
<span class="lineno">3</span>
|
||||
<span class="lineno">4</span><span class="kn">from</span> <span class="nn">labml_helpers.module</span> <span class="kn">import</span> <span class="n">Module</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-1'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-1'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">7</span><span class="k">class</span> <span class="nc">Swish</span><span class="p">(</span><span class="n">Module</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-2'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-2'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">8</span> <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
|
||||
<span class="lineno">9</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
|
||||
<span class="lineno">10</span> <span class="bp">self</span><span class="o">.</span><span class="n">sigmoid</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Sigmoid</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-3'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-3'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">12</span> <span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">)</span> <span class="o">-></span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">:</span>
|
||||
<span class="lineno">13</span> <span class="k">return</span> <span class="n">x</span> <span class="o">*</span> <span class="bp">self</span><span class="o">.</span><span class="n">sigmoid</span><span class="p">(</span><span class="n">x</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
@@ -0,0 +1,128 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content="一组与自适应计算相关的 PyTorch 实现/教程"/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="具有自适应计算的神经网络"/>
|
||||
<meta name="twitter:description" content="一组与自适应计算相关的 PyTorch 实现/教程"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/adaptive_computation/index.html"/>
|
||||
<meta property="og:title" content="具有自适应计算的神经网络"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="具有自适应计算的神经网络"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="具有自适应计算的神经网络"/>
|
||||
<meta property="og:description" content="一组与自适应计算相关的 PyTorch 实现/教程"/>
|
||||
|
||||
<title>具有自适应计算的神经网络</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/adaptive_computation/index.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="index.html">adaptive_computation</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/adaptive_computation/__init__.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1>具有自适应计算的神经网络</h1>
|
||||
<p>这些是神经网络架构，根据输入样本的复杂度来改变计算复杂度。</p>
|
||||
<ul><li>🚧 TODO：循环神经网络的自适应计算时间</li>
|
||||
<li><a href="ponder_net/index.html">PonderNet：学会思考</a></li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
@@ -0,0 +1,276 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content="这将从论文《循环神经网络的自适应计算时间》中为奇偶校验任务创建数据"/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="奇偶校验任务"/>
|
||||
<meta name="twitter:description" content="这将从论文《循环神经网络的自适应计算时间》中为奇偶校验任务创建数据"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/adaptive_computation/parity.html"/>
|
||||
<meta property="og:title" content="奇偶校验任务"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="奇偶校验任务"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="奇偶校验任务"/>
|
||||
<meta property="og:description" content="这将从论文《循环神经网络的自适应计算时间》中为奇偶校验任务创建数据"/>
|
||||
|
||||
<title>奇偶校验任务</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/adaptive_computation/parity.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="index.html">adaptive_computation</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/adaptive_computation/parity.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1>奇偶校验任务</h1>
|
||||
<p>这将从论文《<a href="https://arxiv.org/abs/1603.08983">循环神经网络的自适应计算时间》中为</a>奇偶校验任务创建数据。</p>
|
||||
<p>奇偶校验任务的输入是一个带有<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord">0</span></span></span></span></span>'s 和<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqc" style=""><span class="mord" style="">1</span></span></span></span></span></span>'s 的向量。输出是<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.72777em;vertical-align:-0.08333em;"></span><span class="mord coloredeq eqa" style=""><span class="mord" style="">−</span><span class="mord" style=""><span class="mord coloredeq eqc" style="">1</span></span></span></span></span></span></span>'s 的<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqc" style=""><span class="mord" style="">1</span></span></span></span></span></span>奇偶校验——如果有，则为 1是的奇数<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqc" style=""><span class="mord" style="">1</span></span></span></span></span></span>，否则为零。输入是通过使矢量中的随机数量的元素为<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqc" style=""><span class="mord" style="">1</span></span></span></span></span></span>或而生成<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.72777em;vertical-align:-0.08333em;"></span><span class="mord coloredeq eqa" style=""><span class="mord" style="">−</span><span class="mord" style=""><span class="mord coloredeq eqc" style="">1</span></span></span></span></span></span></span>的。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">19</span><span></span><span class="kn">from</span> <span class="nn">typing</span> <span class="kn">import</span> <span class="n">Tuple</span>
|
||||
<span class="lineno">20</span>
|
||||
<span class="lineno">21</span><span class="kn">import</span> <span class="nn">torch</span>
|
||||
<span class="lineno">22</span><span class="kn">from</span> <span class="nn">torch.utils.data</span> <span class="kn">import</span> <span class="n">Dataset</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-1'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-1'>#</a>
|
||||
</div>
|
||||
<h3>奇偶校验数据</h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">25</span><span class="k">class</span> <span class="nc">ParityDataset</span><span class="p">(</span><span class="n">Dataset</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-2'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-2'>#</a>
|
||||
</div>
|
||||
<ul><li><code class="highlight"><span></span><span class="n">n_samples</span></code>
|
||||
是样本的数量</li>
|
||||
<li><code class="highlight"><span></span><span class="n">n_elems</span></code>
|
||||
是输入向量中的元素数</li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">30</span> <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">n_samples</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">n_elems</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">64</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-3'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-3'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">35</span> <span class="bp">self</span><span class="o">.</span><span class="n">n_samples</span> <span class="o">=</span> <span class="n">n_samples</span>
|
||||
<span class="lineno">36</span> <span class="bp">self</span><span class="o">.</span><span class="n">n_elems</span> <span class="o">=</span> <span class="n">n_elems</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-4'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-4'>#</a>
|
||||
</div>
|
||||
<p>数据集的大小</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">38</span> <span class="k">def</span> <span class="fm">__len__</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-5'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-5'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">42</span> <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">n_samples</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-6'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-6'>#</a>
|
||||
</div>
|
||||
<p>生成样本</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">44</span> <span class="k">def</span> <span class="fm">__getitem__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">idx</span><span class="p">:</span> <span class="nb">int</span><span class="p">)</span> <span class="o">-></span> <span class="n">Tuple</span><span class="p">[</span><span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">]:</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-7'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-7'>#</a>
|
||||
</div>
|
||||
<p>空向量</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">50</span> <span class="n">x</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">zeros</span><span class="p">((</span><span class="bp">self</span><span class="o">.</span><span class="n">n_elems</span><span class="p">,))</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-8'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-8'>#</a>
|
||||
</div>
|
||||
<p>非零元素的数量-介于<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqc" style=""><span class="mord" style="">1</span></span></span></span></span></span>和元素总数之间的随机数</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">52</span> <span class="n">n_non_zero</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">randint</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">n_elems</span> <span class="o">+</span> <span class="mi">1</span><span class="p">,</span> <span class="p">(</span><span class="mi">1</span><span class="p">,))</span><span class="o">.</span><span class="n">item</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-9'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-9'>#</a>
|
||||
</div>
|
||||
<p>用<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqc" style=""><span class="mord" style="">1</span></span></span></span></span></span> “和<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.72777em;vertical-align:-0.08333em;"></span><span class="mord coloredeq eqa" style=""><span class="mord" style="">−</span><span class="mord" style=""><span class="mord coloredeq eqc" style="">1</span></span></span></span></span></span></span>” 填充非零元素</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">54</span> <span class="n">x</span><span class="p">[:</span><span class="n">n_non_zero</span><span class="p">]</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">randint</span><span class="p">(</span><span class="mi">0</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="p">(</span><span class="n">n_non_zero</span><span class="p">,))</span> <span class="o">*</span> <span class="mi">2</span> <span class="o">-</span> <span class="mi">1</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-10'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-10'>#</a>
|
||||
</div>
|
||||
<p>随机排列元素</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">56</span> <span class="n">x</span> <span class="o">=</span> <span class="n">x</span><span class="p">[</span><span class="n">torch</span><span class="o">.</span><span class="n">randperm</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">n_elems</span><span class="p">)]</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-11'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-11'>#</a>
|
||||
</div>
|
||||
<p>奇偶校验</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">59</span> <span class="n">y</span> <span class="o">=</span> <span class="p">(</span><span class="n">x</span> <span class="o">==</span> <span class="mf">1.</span><span class="p">)</span><span class="o">.</span><span class="n">sum</span><span class="p">()</span> <span class="o">%</span> <span class="mi">2</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-12'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-12'>#</a>
|
||||
</div>
|
||||
<p></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">62</span> <span class="k">return</span> <span class="n">x</span><span class="p">,</span> <span class="n">y</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
@@ -0,0 +1,639 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content="这列火车是奇偶校验任务上的 PonderNet"/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="PonderNet 奇偶校验任务实验"/>
|
||||
<meta name="twitter:description" content="这列火车是奇偶校验任务上的 PonderNet"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/adaptive_computation/ponder_net/experiment.html"/>
|
||||
<meta property="og:title" content="PonderNet 奇偶校验任务实验"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="PonderNet 奇偶校验任务实验"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="PonderNet 奇偶校验任务实验"/>
|
||||
<meta property="og:description" content="这列火车是奇偶校验任务上的 PonderNet"/>
|
||||
|
||||
<title>PonderNet 奇偶校验任务实验</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/adaptive_computation/ponder_net/experiment.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="../index.html">adaptive_computation</a>
|
||||
<a class="parent" href="index.html">ponder_net</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/adaptive_computation/ponder_net/experiment.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1><a href="index.html">PonderNet</a> <a href="../parity.html">奇偶校验任务</a>实验</h1>
|
||||
<p>这会在<a href="../parity.html">奇偶校验任务</a>上训练 <a href="index.html">PonderNet</a>。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">13</span><span></span><span class="kn">from</span> <span class="nn">typing</span> <span class="kn">import</span> <span class="n">Any</span>
|
||||
<span class="lineno">14</span>
|
||||
<span class="lineno">15</span><span class="kn">import</span> <span class="nn">torch</span>
|
||||
<span class="lineno">16</span><span class="kn">from</span> <span class="nn">torch</span> <span class="kn">import</span> <span class="n">nn</span>
|
||||
<span class="lineno">17</span><span class="kn">from</span> <span class="nn">torch.utils.data</span> <span class="kn">import</span> <span class="n">DataLoader</span>
|
||||
<span class="lineno">18</span>
|
||||
<span class="lineno">19</span><span class="kn">from</span> <span class="nn">labml</span> <span class="kn">import</span> <span class="n">tracker</span><span class="p">,</span> <span class="n">experiment</span>
|
||||
<span class="lineno">20</span><span class="kn">from</span> <span class="nn">labml_helpers.metrics.accuracy</span> <span class="kn">import</span> <span class="n">AccuracyDirect</span>
|
||||
<span class="lineno">21</span><span class="kn">from</span> <span class="nn">labml_helpers.train_valid</span> <span class="kn">import</span> <span class="n">SimpleTrainValidConfigs</span><span class="p">,</span> <span class="n">BatchIndex</span>
|
||||
<span class="lineno">22</span><span class="kn">from</span> <span class="nn">labml_nn.adaptive_computation.parity</span> <span class="kn">import</span> <span class="n">ParityDataset</span>
|
||||
<span class="lineno">23</span><span class="kn">from</span> <span class="nn">labml_nn.adaptive_computation.ponder_net</span> <span class="kn">import</span> <span class="n">ParityPonderGRU</span><span class="p">,</span> <span class="n">ReconstructionLoss</span><span class="p">,</span> <span class="n">RegularizationLoss</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-1'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-1'>#</a>
|
||||
</div>
|
||||
<p>带有<a href="https://docs.labml.ai/api/helpers.html#labml_helpers.train_valid.SimpleTrainValidConfigs">简单训练循环</a>的配置</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">26</span><span class="k">class</span> <span class="nc">Configs</span><span class="p">(</span><span class="n">SimpleTrainValidConfigs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-2'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-2'>#</a>
|
||||
</div>
|
||||
<p>周期的数量</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">33</span> <span class="n">epochs</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">100</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-3'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-3'>#</a>
|
||||
</div>
|
||||
<p>每个纪元的批次数</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">35</span> <span class="n">n_batches</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">500</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-4'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-4'>#</a>
|
||||
</div>
|
||||
<p>批量大小</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">37</span> <span class="n">batch_size</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">128</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-5'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-5'>#</a>
|
||||
</div>
|
||||
<p>型号</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">40</span> <span class="n">model</span><span class="p">:</span> <span class="n">ParityPonderGRU</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-6'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-6'>#</a>
|
||||
</div>
|
||||
<p><span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.83333em;vertical-align:-0.15em;"></span><span class="mord coloredeq eqd" style=""><span class="mord" style=""><span class="mord mathnormal" style="">L</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.32833099999999993em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style=""><span class="mord mathnormal mtight" style="margin-right:0.00773em">R</span><span class="mord mathnormal mtight" style="">ec</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">43</span> <span class="n">loss_rec</span><span class="p">:</span> <span class="n">ReconstructionLoss</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-7'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-7'>#</a>
|
||||
</div>
|
||||
<p><span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.969438em;vertical-align:-0.286108em;"></span><span class="mord coloredeq eqe" style=""><span class="mord" style=""><span class="mord mathnormal" style="">L</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.328331em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style=""><span class="mord mathnormal mtight" style="margin-right:0.00773em">R</span><span class="mord mathnormal mtight" style="">e</span><span class="mord mathnormal mtight" style="margin-right:0.03588em">g</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.286108em;"><span></span></span></span></span></span></span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">45</span> <span class="n">loss_reg</span><span class="p">:</span> <span class="n">RegularizationLoss</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-8'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-8'>#</a>
|
||||
</div>
|
||||
<p>输入向量中的元素数。<em>我们将其保持在较低的水平以进行演示；否则，训练会花费很多时间。尽管奇偶校验任务看起来很简单，但通过查看样本来找出模式相当困难。</em></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">51</span> <span class="n">n_elems</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">8</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-9'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-9'>#</a>
|
||||
</div>
|
||||
<p>隐藏层（状态）中的单位数量</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">53</span> <span class="n">n_hidden</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">64</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-10'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-10'>#</a>
|
||||
</div>
|
||||
<p>最大步数<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.68333em;vertical-align:0em;"></span><span class="mord mathnormal" style="margin-right:0.10903em;">N</span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">55</span> <span class="n">max_steps</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">20</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-11'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-11'>#</a>
|
||||
</div>
|
||||
<p><span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.980548em;vertical-align:-0.286108em;"></span><span class="mord coloredeq eqc" style=""><span class="mord" style=""><span class="mord mathnormal" style="">λ</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.15139200000000003em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mathnormal mtight" style="">p</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.286108em;"><span></span></span></span></span></span></span></span></span></span></span></span>用于几何分布<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:1.036108em;vertical-align:-0.286108em;"></span><span class="mord"><span class="mord mathnormal">p</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.32833099999999993em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight"><span class="mord mathnormal mtight">G</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span><span class="mopen">(</span><span class="mord coloredeq eqc" style=""><span class="mord" style=""><span class="mord mathnormal" style="">λ</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.15139200000000003em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mathnormal mtight" style="">p</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.286108em;"><span></span></span></span></span></span></span></span><span class="mclose">)</span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">58</span> <span class="n">lambda_p</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">0.2</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-12'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-12'>#</a>
|
||||
</div>
|
||||
<p>正则化损失<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.969438em;vertical-align:-0.286108em;"></span><span class="mord coloredeq eqe" style=""><span class="mord" style=""><span class="mord mathnormal" style="">L</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.328331em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style=""><span class="mord mathnormal mtight" style="margin-right:0.00773em">R</span><span class="mord mathnormal mtight" style="">e</span><span class="mord mathnormal mtight" style="margin-right:0.03588em">g</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.286108em;"><span></span></span></span></span></span></span></span></span></span></span></span>系数<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.8888799999999999em;vertical-align:-0.19444em;"></span><span class="mord coloredeq eqf" style=""><span class="mord mathnormal" style="margin-right:0.05278em">β</span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">60</span> <span class="n">beta</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">0.01</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-13'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-13'>#</a>
|
||||
</div>
|
||||
<p>按规范进行渐变裁剪</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">63</span> <span class="n">grad_norm_clip</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">1.0</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-14'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-14'>#</a>
|
||||
</div>
|
||||
<p>训练和验证装载机</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">66</span> <span class="n">train_loader</span><span class="p">:</span> <span class="n">DataLoader</span>
|
||||
<span class="lineno">67</span> <span class="n">valid_loader</span><span class="p">:</span> <span class="n">DataLoader</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-15'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-15'>#</a>
|
||||
</div>
|
||||
<p>精度计算器</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">70</span> <span class="n">accuracy</span> <span class="o">=</span> <span class="n">AccuracyDirect</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-16'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-16'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">72</span> <span class="k">def</span> <span class="nf">init</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-17'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-17'>#</a>
|
||||
</div>
|
||||
<p>将指示器打印到屏幕上</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">74</span> <span class="n">tracker</span><span class="o">.</span><span class="n">set_scalar</span><span class="p">(</span><span class="s1">'loss.*'</span><span class="p">,</span> <span class="kc">True</span><span class="p">)</span>
|
||||
<span class="lineno">75</span> <span class="n">tracker</span><span class="o">.</span><span class="n">set_scalar</span><span class="p">(</span><span class="s1">'loss_reg.*'</span><span class="p">,</span> <span class="kc">True</span><span class="p">)</span>
|
||||
<span class="lineno">76</span> <span class="n">tracker</span><span class="o">.</span><span class="n">set_scalar</span><span class="p">(</span><span class="s1">'accuracy.*'</span><span class="p">,</span> <span class="kc">True</span><span class="p">)</span>
|
||||
<span class="lineno">77</span> <span class="n">tracker</span><span class="o">.</span><span class="n">set_scalar</span><span class="p">(</span><span class="s1">'steps.*'</span><span class="p">,</span> <span class="kc">True</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-18'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-18'>#</a>
|
||||
</div>
|
||||
<p>我们需要设置指标来计算训练和验证时期的指标</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">80</span> <span class="bp">self</span><span class="o">.</span><span class="n">state_modules</span> <span class="o">=</span> <span class="p">[</span><span class="bp">self</span><span class="o">.</span><span class="n">accuracy</span><span class="p">]</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-19'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-19'>#</a>
|
||||
</div>
|
||||
<p>初始化模型</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">83</span> <span class="bp">self</span><span class="o">.</span><span class="n">model</span> <span class="o">=</span> <span class="n">ParityPonderGRU</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">n_elems</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">n_hidden</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">max_steps</span><span class="p">)</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">device</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-20'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-20'>#</a>
|
||||
</div>
|
||||
<p><span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.83333em;vertical-align:-0.15em;"></span><span class="mord coloredeq eqd" style=""><span class="mord" style=""><span class="mord mathnormal" style="">L</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.32833099999999993em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style=""><span class="mord mathnormal mtight" style="margin-right:0.00773em">R</span><span class="mord mathnormal mtight" style="">ec</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">85</span> <span class="bp">self</span><span class="o">.</span><span class="n">loss_rec</span> <span class="o">=</span> <span class="n">ReconstructionLoss</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">BCEWithLogitsLoss</span><span class="p">(</span><span class="n">reduction</span><span class="o">=</span><span class="s1">'none'</span><span class="p">))</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">device</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-21'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-21'>#</a>
|
||||
</div>
|
||||
<p><span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.969438em;vertical-align:-0.286108em;"></span><span class="mord coloredeq eqe" style=""><span class="mord" style=""><span class="mord mathnormal" style="">L</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.328331em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style=""><span class="mord mathnormal mtight" style="margin-right:0.00773em">R</span><span class="mord mathnormal mtight" style="">e</span><span class="mord mathnormal mtight" style="margin-right:0.03588em">g</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.286108em;"><span></span></span></span></span></span></span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">87</span> <span class="bp">self</span><span class="o">.</span><span class="n">loss_reg</span> <span class="o">=</span> <span class="n">RegularizationLoss</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">lambda_p</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">max_steps</span><span class="p">)</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">device</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-22'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-22'>#</a>
|
||||
</div>
|
||||
<p>训练和验证装载机</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">90</span> <span class="bp">self</span><span class="o">.</span><span class="n">train_loader</span> <span class="o">=</span> <span class="n">DataLoader</span><span class="p">(</span><span class="n">ParityDataset</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">batch_size</span> <span class="o">*</span> <span class="bp">self</span><span class="o">.</span><span class="n">n_batches</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">n_elems</span><span class="p">),</span>
|
||||
<span class="lineno">91</span> <span class="n">batch_size</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">batch_size</span><span class="p">)</span>
|
||||
<span class="lineno">92</span> <span class="bp">self</span><span class="o">.</span><span class="n">valid_loader</span> <span class="o">=</span> <span class="n">DataLoader</span><span class="p">(</span><span class="n">ParityDataset</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">batch_size</span> <span class="o">*</span> <span class="mi">32</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">n_elems</span><span class="p">),</span>
|
||||
<span class="lineno">93</span> <span class="n">batch_size</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">batch_size</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-23'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-23'>#</a>
|
||||
</div>
|
||||
<p>培训师会为每批次调用此方法</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">95</span> <span class="k">def</span> <span class="nf">step</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">batch</span><span class="p">:</span> <span class="n">Any</span><span class="p">,</span> <span class="n">batch_idx</span><span class="p">:</span> <span class="n">BatchIndex</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-24'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-24'>#</a>
|
||||
</div>
|
||||
<p>设置模型模式</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">100</span> <span class="bp">self</span><span class="o">.</span><span class="n">model</span><span class="o">.</span><span class="n">train</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">mode</span><span class="o">.</span><span class="n">is_train</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-25'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-25'>#</a>
|
||||
</div>
|
||||
<p>获取输入和标签并将其移动到模型的设备中</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">103</span> <span class="n">data</span><span class="p">,</span> <span class="n">target</span> <span class="o">=</span> <span class="n">batch</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">device</span><span class="p">),</span> <span class="n">batch</span><span class="p">[</span><span class="mi">1</span><span class="p">]</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">device</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-26'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-26'>#</a>
|
||||
</div>
|
||||
<p>在训练模式中增加步数</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">106</span> <span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">mode</span><span class="o">.</span><span class="n">is_train</span><span class="p">:</span>
|
||||
<span class="lineno">107</span> <span class="n">tracker</span><span class="o">.</span><span class="n">add_global_step</span><span class="p">(</span><span class="nb">len</span><span class="p">(</span><span class="n">data</span><span class="p">))</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-27'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-27'>#</a>
|
||||
</div>
|
||||
<p>运行模型</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">110</span> <span class="n">p</span><span class="p">,</span> <span class="n">y_hat</span><span class="p">,</span> <span class="n">p_sampled</span><span class="p">,</span> <span class="n">y_hat_sampled</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">model</span><span class="p">(</span><span class="n">data</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-28'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-28'>#</a>
|
||||
</div>
|
||||
<p>计算重建损失</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">113</span> <span class="n">loss_rec</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">loss_rec</span><span class="p">(</span><span class="n">p</span><span class="p">,</span> <span class="n">y_hat</span><span class="p">,</span> <span class="n">target</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">float</span><span class="p">))</span>
|
||||
<span class="lineno">114</span> <span class="n">tracker</span><span class="o">.</span><span class="n">add</span><span class="p">(</span><span class="s2">"loss."</span><span class="p">,</span> <span class="n">loss_rec</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-29'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-29'>#</a>
|
||||
</div>
|
||||
<p>计算正则化损失</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">117</span> <span class="n">loss_reg</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">loss_reg</span><span class="p">(</span><span class="n">p</span><span class="p">)</span>
|
||||
<span class="lineno">118</span> <span class="n">tracker</span><span class="o">.</span><span class="n">add</span><span class="p">(</span><span class="s2">"loss_reg."</span><span class="p">,</span> <span class="n">loss_reg</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-30'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-30'>#</a>
|
||||
</div>
|
||||
<p><span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.68333em;vertical-align:0em;"></span><span class="mord mathnormal">L</span><span class="mspace" style="margin-right:0.2777777777777778em;"></span><span class="mrel">=</span><span class="mspace" style="margin-right:0.2777777777777778em;"></span></span><span class="base"><span class="strut" style="height:0.83333em;vertical-align:-0.15em;"></span><span class="mord coloredeq eqd" style=""><span class="mord" style=""><span class="mord mathnormal" style="">L</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.32833099999999993em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style=""><span class="mord mathnormal mtight" style="margin-right:0.00773em">R</span><span class="mord mathnormal mtight" style="">ec</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span></span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin">+</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span></span><span class="base"><span class="strut" style="height:0.980548em;vertical-align:-0.286108em;"></span><span class="mord coloredeq eqf" style=""><span class="mord mathnormal" style="margin-right:0.05278em">β</span></span><span class="mord coloredeq eqe" style=""><span class="mord" style=""><span class="mord mathnormal" style="">L</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.328331em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style=""><span class="mord mathnormal mtight" style="margin-right:0.00773em">R</span><span class="mord mathnormal mtight" style="">e</span><span class="mord mathnormal mtight" style="margin-right:0.03588em">g</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.286108em;"><span></span></span></span></span></span></span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">121</span> <span class="n">loss</span> <span class="o">=</span> <span class="n">loss_rec</span> <span class="o">+</span> <span class="bp">self</span><span class="o">.</span><span class="n">beta</span> <span class="o">*</span> <span class="n">loss_reg</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-31'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-31'>#</a>
|
||||
</div>
|
||||
<p>计算预期采取的步数</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">124</span> <span class="n">steps</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">arange</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="n">p</span><span class="o">.</span><span class="n">shape</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span> <span class="o">+</span> <span class="mi">1</span><span class="p">,</span> <span class="n">device</span><span class="o">=</span><span class="n">p</span><span class="o">.</span><span class="n">device</span><span class="p">)</span>
|
||||
<span class="lineno">125</span> <span class="n">expected_steps</span> <span class="o">=</span> <span class="p">(</span><span class="n">p</span> <span class="o">*</span> <span class="n">steps</span><span class="p">[:,</span> <span class="kc">None</span><span class="p">])</span><span class="o">.</span><span class="n">sum</span><span class="p">(</span><span class="n">dim</span><span class="o">=</span><span class="mi">0</span><span class="p">)</span>
|
||||
<span class="lineno">126</span> <span class="n">tracker</span><span class="o">.</span><span class="n">add</span><span class="p">(</span><span class="s2">"steps."</span><span class="p">,</span> <span class="n">expected_steps</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-32'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-32'>#</a>
|
||||
</div>
|
||||
<p>呼叫准确度指标</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">129</span> <span class="bp">self</span><span class="o">.</span><span class="n">accuracy</span><span class="p">(</span><span class="n">y_hat_sampled</span> <span class="o">></span> <span class="mi">0</span><span class="p">,</span> <span class="n">target</span><span class="p">)</span>
|
||||
<span class="lineno">130</span>
|
||||
<span class="lineno">131</span> <span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">mode</span><span class="o">.</span><span class="n">is_train</span><span class="p">:</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-33'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-33'>#</a>
|
||||
</div>
|
||||
<p>计算梯度</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">133</span> <span class="n">loss</span><span class="o">.</span><span class="n">backward</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-34'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-34'>#</a>
|
||||
</div>
|
||||
<p>剪辑渐变</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">135</span> <span class="n">torch</span><span class="o">.</span><span class="n">nn</span><span class="o">.</span><span class="n">utils</span><span class="o">.</span><span class="n">clip_grad_norm_</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">model</span><span class="o">.</span><span class="n">parameters</span><span class="p">(),</span> <span class="n">max_norm</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">grad_norm_clip</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-35'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-35'>#</a>
|
||||
</div>
|
||||
<p>优化器</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">137</span> <span class="bp">self</span><span class="o">.</span><span class="n">optimizer</span><span class="o">.</span><span class="n">step</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-36'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-36'>#</a>
|
||||
</div>
|
||||
<p>渐变清晰</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">139</span> <span class="bp">self</span><span class="o">.</span><span class="n">optimizer</span><span class="o">.</span><span class="n">zero_grad</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-37'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-37'>#</a>
|
||||
</div>
|
||||
<p></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">141</span> <span class="n">tracker</span><span class="o">.</span><span class="n">save</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-38'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-38'>#</a>
|
||||
</div>
|
||||
<p>运行实验</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">144</span><span class="k">def</span> <span class="nf">main</span><span class="p">():</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-39'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-39'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">148</span> <span class="n">experiment</span><span class="o">.</span><span class="n">create</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">'ponder_net'</span><span class="p">)</span>
|
||||
<span class="lineno">149</span>
|
||||
<span class="lineno">150</span> <span class="n">conf</span> <span class="o">=</span> <span class="n">Configs</span><span class="p">()</span>
|
||||
<span class="lineno">151</span> <span class="n">experiment</span><span class="o">.</span><span class="n">configs</span><span class="p">(</span><span class="n">conf</span><span class="p">,</span> <span class="p">{</span>
|
||||
<span class="lineno">152</span> <span class="s1">'optimizer.optimizer'</span><span class="p">:</span> <span class="s1">'Adam'</span><span class="p">,</span>
|
||||
<span class="lineno">153</span> <span class="s1">'optimizer.learning_rate'</span><span class="p">:</span> <span class="mf">0.0003</span><span class="p">,</span>
|
||||
<span class="lineno">154</span> <span class="p">})</span>
|
||||
<span class="lineno">155</span>
|
||||
<span class="lineno">156</span> <span class="k">with</span> <span class="n">experiment</span><span class="o">.</span><span class="n">start</span><span class="p">():</span>
|
||||
<span class="lineno">157</span> <span class="n">conf</span><span class="o">.</span><span class="n">run</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-40'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-40'>#</a>
|
||||
</div>
|
||||
<p></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">160</span><span class="k">if</span> <span class="vm">__name__</span> <span class="o">==</span> <span class="s1">'__main__'</span><span class="p">:</span>
|
||||
<span class="lineno">161</span> <span class="n">main</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
File diff suppressed because one or more lines are too long
@@ -0,0 +1,128 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content=""/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="PonderNet：学会思考"/>
|
||||
<meta name="twitter:description" content=""/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/adaptive_computation/ponder_net/readme.html"/>
|
||||
<meta property="og:title" content="PonderNet：学会思考"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="PonderNet：学会思考"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="PonderNet：学会思考"/>
|
||||
<meta property="og:description" content=""/>
|
||||
|
||||
<title>PonderNet：学会思考</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/adaptive_computation/ponder_net/readme.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="../index.html">adaptive_computation</a>
|
||||
<a class="parent" href="index.html">ponder_net</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/adaptive_computation/ponder_net/readme.md" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1><a href="https://nn.labml.ai/adaptive_computation/ponder_net/index.html">PonderNet：学会思考</a></h1>
|
||||
<p>这是 <a href="https://pytorch.org">P <a href="https://arxiv.org/abs/2107.05407">onderNet：学会思考</a>论文的 PyTorch</a> 实现。</p>
|
||||
<p>PonderNet 根据输入调整计算。它会根据输入更改循环网络上要执行的步骤数。PonderNet 通过端到端梯度下降来学习这一点。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
@@ -0,0 +1,128 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content=""/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="具有自适应计算的神经网络"/>
|
||||
<meta name="twitter:description" content=""/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/adaptive_computation/readme.html"/>
|
||||
<meta property="og:title" content="具有自适应计算的神经网络"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="具有自适应计算的神经网络"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="具有自适应计算的神经网络"/>
|
||||
<meta property="og:description" content=""/>
|
||||
|
||||
<title>具有自适应计算的神经网络</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/adaptive_computation/readme.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="index.html">adaptive_computation</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/adaptive_computation/readme.md" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1><a href="https://nn.labml.ai/adaptive_computation/index.html">具有自适应计算的神经网络</a></h1>
|
||||
<p>这些是神经网络架构，根据输入样本的复杂度来改变计算复杂度。</p>
|
||||
<ul><li>🚧 TODO：循环神经网络的自适应计算时间</li>
|
||||
<li><a href="https://nn.labml.ai/adaptive_computation/ponder_net/index.html">PonderNet：学会思考</a></li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
File diff suppressed because one or more lines are too long
@@ -0,0 +1,599 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content="在 MNIST 数据集上训练胶囊网络的代码"/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="使用胶囊网络对 MNIST 数字进行分类"/>
|
||||
<meta name="twitter:description" content="在 MNIST 数据集上训练胶囊网络的代码"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/capsule_networks/mnist.html"/>
|
||||
<meta property="og:title" content="使用胶囊网络对 MNIST 数字进行分类"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="使用胶囊网络对 MNIST 数字进行分类"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="使用胶囊网络对 MNIST 数字进行分类"/>
|
||||
<meta property="og:description" content="在 MNIST 数据集上训练胶囊网络的代码"/>
|
||||
|
||||
<title>使用胶囊网络对 MNIST 数字进行分类</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/capsule_networks/mnist.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="index.html">capsule_networks</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/capsule_networks/mnist.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1>使用胶囊网络对 MNIST 数字进行分类</h1>
|
||||
<p>这是一个带注释的 PyTorch 代码，用于使用 PyTorch 对 MNIST 数字进行分类。</p>
|
||||
<p>本文实施了论文《<a href="https://arxiv.org/abs/1710.09829">胶囊间动态路由</a>》中描述的实验。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">14</span><span></span><span class="kn">from</span> <span class="nn">typing</span> <span class="kn">import</span> <span class="n">Any</span>
|
||||
<span class="lineno">15</span>
|
||||
<span class="lineno">16</span><span class="kn">import</span> <span class="nn">torch.nn</span> <span class="k">as</span> <span class="nn">nn</span>
|
||||
<span class="lineno">17</span><span class="kn">import</span> <span class="nn">torch.nn.functional</span> <span class="k">as</span> <span class="nn">F</span>
|
||||
<span class="lineno">18</span><span class="kn">import</span> <span class="nn">torch.utils.data</span>
|
||||
<span class="lineno">19</span>
|
||||
<span class="lineno">20</span><span class="kn">from</span> <span class="nn">labml</span> <span class="kn">import</span> <span class="n">experiment</span><span class="p">,</span> <span class="n">tracker</span>
|
||||
<span class="lineno">21</span><span class="kn">from</span> <span class="nn">labml.configs</span> <span class="kn">import</span> <span class="n">option</span>
|
||||
<span class="lineno">22</span><span class="kn">from</span> <span class="nn">labml_helpers.datasets.mnist</span> <span class="kn">import</span> <span class="n">MNISTConfigs</span>
|
||||
<span class="lineno">23</span><span class="kn">from</span> <span class="nn">labml_helpers.metrics.accuracy</span> <span class="kn">import</span> <span class="n">AccuracyDirect</span>
|
||||
<span class="lineno">24</span><span class="kn">from</span> <span class="nn">labml_helpers.module</span> <span class="kn">import</span> <span class="n">Module</span>
|
||||
<span class="lineno">25</span><span class="kn">from</span> <span class="nn">labml_helpers.train_valid</span> <span class="kn">import</span> <span class="n">SimpleTrainValidConfigs</span><span class="p">,</span> <span class="n">BatchIndex</span>
|
||||
<span class="lineno">26</span><span class="kn">from</span> <span class="nn">labml_nn.capsule_networks</span> <span class="kn">import</span> <span class="n">Squash</span><span class="p">,</span> <span class="n">Router</span><span class="p">,</span> <span class="n">MarginLoss</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-1'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-1'>#</a>
|
||||
</div>
|
||||
<h2>用于对 MNIST 数字进行分类的模型</h2>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">29</span><span class="k">class</span> <span class="nc">MNISTCapsuleNetworkModel</span><span class="p">(</span><span class="n">Module</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-2'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-2'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">34</span> <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
|
||||
<span class="lineno">35</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-3'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-3'>#</a>
|
||||
</div>
|
||||
<p>第一个卷积层有<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eql" style=""><span class="mord" style="">2</span></span><span class="mord">56</span></span></span></span></span>，<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.72777em;vertical-align:-0.08333em;"></span><span class="mord">9</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin">×</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span></span><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord">9</span></span></span></span></span>卷积内核</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">37</span> <span class="bp">self</span><span class="o">.</span><span class="n">conv1</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Conv2d</span><span class="p">(</span><span class="n">in_channels</span><span class="o">=</span><span class="mi">1</span><span class="p">,</span> <span class="n">out_channels</span><span class="o">=</span><span class="mi">256</span><span class="p">,</span> <span class="n">kernel_size</span><span class="o">=</span><span class="mi">9</span><span class="p">,</span> <span class="n">stride</span><span class="o">=</span><span class="mi">1</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-4'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-4'>#</a>
|
||||
</div>
|
||||
<p>第二层（Primary Capsules）是卷积胶囊层，带有卷积<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.68333em;vertical-align:0em;"></span><span class="mord coloredeq eqn" style=""><span class="mord" style="">8</span></span><span class="mord mathnormal" style="margin-right:0.02778em;">D</span></span></span></span></span>胶囊<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqj" style=""><span class="mord" style=""><span class="mord coloredeq eqm" style="">3</span></span><span class="mord" style=""><span class="mord coloredeq eql" style="">2</span></span></span></span></span></span></span>通道（每个胶囊<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqn" style=""><span class="mord" style="">8</span></span></span></span></span></span>的特征）。也就是说，每个主胶囊包含 8 个卷积单位，内核为 9×9，步幅为 2。为了实现这一点，我们创建了一个带有<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.72777em;vertical-align:-0.08333em;"></span><span class="mord coloredeq eqj" style=""><span class="mord" style=""><span class="mord coloredeq eqm" style="">3</span></span><span class="mord" style=""><span class="mord coloredeq eql" style="">2</span></span></span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin">×</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span></span><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqn" style=""><span class="mord" style="">8</span></span></span></span></span></span>通道的卷积层，并对其输出进行整形和排列，以获得每个<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqn" style=""><span class="mord" style="">8</span></span></span></span></span></span>特征的胶囊。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">43</span> <span class="bp">self</span><span class="o">.</span><span class="n">conv2</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Conv2d</span><span class="p">(</span><span class="n">in_channels</span><span class="o">=</span><span class="mi">256</span><span class="p">,</span> <span class="n">out_channels</span><span class="o">=</span><span class="mi">32</span> <span class="o">*</span> <span class="mi">8</span><span class="p">,</span> <span class="n">kernel_size</span><span class="o">=</span><span class="mi">9</span><span class="p">,</span> <span class="n">stride</span><span class="o">=</span><span class="mi">2</span><span class="p">,</span> <span class="n">padding</span><span class="o">=</span><span class="mi">0</span><span class="p">)</span>
|
||||
<span class="lineno">44</span> <span class="bp">self</span><span class="o">.</span><span class="n">squash</span> <span class="o">=</span> <span class="n">Squash</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-5'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-5'>#</a>
|
||||
</div>
|
||||
<p>路由层获取<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.72777em;vertical-align:-0.08333em;"></span><span class="mord coloredeq eqj" style=""><span class="mord" style=""><span class="mord coloredeq eqm" style="">3</span></span><span class="mord" style=""><span class="mord coloredeq eql" style="">2</span></span></span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin">×</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span></span><span class="base"><span class="strut" style="height:0.72777em;vertical-align:-0.08333em;"></span><span class="mord">6</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin">×</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span></span><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord">6</span></span></span></span></span>主胶囊并生成<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqh" style=""><span class="mord" style="">10</span></span></span></span></span></span>胶囊。每个主胶囊都有<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqn" style=""><span class="mord" style="">8</span></span></span></span></span></span>特征，而输出胶囊（Digit Capsules）都有<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqi" style=""><span class="mord" style="">16</span></span></span></span></span></span>特征。路由算法会迭代<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqm" style=""><span class="mord" style="">3</span></span></span></span></span></span>次数。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">50</span> <span class="bp">self</span><span class="o">.</span><span class="n">digit_capsules</span> <span class="o">=</span> <span class="n">Router</span><span class="p">(</span><span class="mi">32</span> <span class="o">*</span> <span class="mi">6</span> <span class="o">*</span> <span class="mi">6</span><span class="p">,</span> <span class="mi">10</span><span class="p">,</span> <span class="mi">8</span><span class="p">,</span> <span class="mi">16</span><span class="p">,</span> <span class="mi">3</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-6'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-6'>#</a>
|
||||
</div>
|
||||
<p>这是本文中提到的解码器。它采用<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqh" style=""><span class="mord" style="">10</span></span></span></span></span></span>数字胶囊的输出，每个胶囊都有重现图像的<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqi" style=""><span class="mord" style="">16</span></span></span></span></span></span>功能。它穿过大小<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord">51</span><span class="mord coloredeq eql" style=""><span class="mord" style="">2</span></span></span></span></span></span>和<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.68333em;vertical-align:0em;"></span><span class="mord mathnormal" style="margin-right:0.00773em;">R</span><span class="mord mathnormal">e</span><span class="mord mathnormal" style="margin-right:0.10903em;">LU</span></span></span></span></span>激活<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqh" style=""><span class="mord" style="">10</span></span><span class="mord coloredeq eql" style=""><span class="mord" style="">2</span></span><span class="mord">4</span></span></span></span></span>的线性层。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">55</span> <span class="bp">self</span><span class="o">.</span><span class="n">decoder</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Sequential</span><span class="p">(</span>
|
||||
<span class="lineno">56</span> <span class="n">nn</span><span class="o">.</span><span class="n">Linear</span><span class="p">(</span><span class="mi">16</span> <span class="o">*</span> <span class="mi">10</span><span class="p">,</span> <span class="mi">512</span><span class="p">),</span>
|
||||
<span class="lineno">57</span> <span class="n">nn</span><span class="o">.</span><span class="n">ReLU</span><span class="p">(),</span>
|
||||
<span class="lineno">58</span> <span class="n">nn</span><span class="o">.</span><span class="n">Linear</span><span class="p">(</span><span class="mi">512</span><span class="p">,</span> <span class="mi">1024</span><span class="p">),</span>
|
||||
<span class="lineno">59</span> <span class="n">nn</span><span class="o">.</span><span class="n">ReLU</span><span class="p">(),</span>
|
||||
<span class="lineno">60</span> <span class="n">nn</span><span class="o">.</span><span class="n">Linear</span><span class="p">(</span><span class="mi">1024</span><span class="p">,</span> <span class="mi">784</span><span class="p">),</span>
|
||||
<span class="lineno">61</span> <span class="n">nn</span><span class="o">.</span><span class="n">Sigmoid</span><span class="p">()</span>
|
||||
<span class="lineno">62</span> <span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-7'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-7'>#</a>
|
||||
</div>
|
||||
<p><code class="highlight"><span></span><span class="n">data</span></code>
|
||||
是 MNIST 图像，有形状<code class="highlight"><span></span><span class="p">[</span><span class="n">batch_size</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">28</span><span class="p">,</span> <span class="mi">28</span><span class="p">]</span></code>
|
||||
</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">64</span> <span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">data</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-8'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-8'>#</a>
|
||||
</div>
|
||||
<p>穿过第一个卷积层。此图层的输出具有形状<code class="highlight"><span></span><span class="p">[</span><span class="n">batch_size</span><span class="p">,</span> <span class="mi">256</span><span class="p">,</span> <span class="mi">20</span><span class="p">,</span> <span class="mi">20</span><span class="p">]</span></code>
|
||||
</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">70</span> <span class="n">x</span> <span class="o">=</span> <span class="n">F</span><span class="o">.</span><span class="n">relu</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">conv1</span><span class="p">(</span><span class="n">data</span><span class="p">))</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-9'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-9'>#</a>
|
||||
</div>
|
||||
<p>穿过第二个卷积层。这个的输出有形状<code class="highlight"><span></span><span class="p">[</span><span class="n">batch_size</span><span class="p">,</span> <span class="mi">32</span> <span class="o">*</span> <span class="mi">8</span><span class="p">,</span> <span class="mi">6</span><span class="p">,</span> <span class="mi">6</span><span class="p">]</span></code>
|
||||
。<em>请注意，此图层的步长为<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eql" style=""><span class="mord" style="">2</span></span></span></span></span></span></em>。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">74</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">conv2</span><span class="p">(</span><span class="n">x</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-10'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-10'>#</a>
|
||||
</div>
|
||||
<p>调整大小并排列以获得胶囊</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">77</span> <span class="n">caps</span> <span class="o">=</span> <span class="n">x</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="n">x</span><span class="o">.</span><span class="n">shape</span><span class="p">[</span><span class="mi">0</span><span class="p">],</span> <span class="mi">8</span><span class="p">,</span> <span class="mi">32</span> <span class="o">*</span> <span class="mi">6</span> <span class="o">*</span> <span class="mi">6</span><span class="p">)</span><span class="o">.</span><span class="n">permute</span><span class="p">(</span><span class="mi">0</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-11'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-11'>#</a>
|
||||
</div>
|
||||
<p>挤压胶囊</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">79</span> <span class="n">caps</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">squash</span><span class="p">(</span><span class="n">caps</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-12'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-12'>#</a>
|
||||
</div>
|
||||
<p>带他们通过路由器获得数字胶囊。这有形状<code class="highlight"><span></span><span class="p">[</span><span class="n">batch_size</span><span class="p">,</span> <span class="mi">10</span><span class="p">,</span> <span class="mi">16</span><span class="p">]</span></code>
|
||||
。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">82</span> <span class="n">caps</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">digit_capsules</span><span class="p">(</span><span class="n">caps</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-13'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-13'>#</a>
|
||||
</div>
|
||||
<p>获取用于重建的口罩</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">85</span> <span class="k">with</span> <span class="n">torch</span><span class="o">.</span><span class="n">no_grad</span><span class="p">():</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-14'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-14'>#</a>
|
||||
</div>
|
||||
<p>胶囊网络的预测是长度最长的胶囊</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">87</span> <span class="n">pred</span> <span class="o">=</span> <span class="p">(</span><span class="n">caps</span> <span class="o">**</span> <span class="mi">2</span><span class="p">)</span><span class="o">.</span><span class="n">sum</span><span class="p">(</span><span class="o">-</span><span class="mi">1</span><span class="p">)</span><span class="o">.</span><span class="n">argmax</span><span class="p">(</span><span class="o">-</span><span class="mi">1</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-15'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-15'>#</a>
|
||||
</div>
|
||||
<p>创建遮罩以遮盖所有其他胶囊</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">89</span> <span class="n">mask</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">eye</span><span class="p">(</span><span class="mi">10</span><span class="p">,</span> <span class="n">device</span><span class="o">=</span><span class="n">data</span><span class="o">.</span><span class="n">device</span><span class="p">)[</span><span class="n">pred</span><span class="p">]</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-16'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-16'>#</a>
|
||||
</div>
|
||||
<p>掩盖数字胶囊以仅获得做出预测的胶囊，然后将其通过解码器进行重建</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">93</span> <span class="n">reconstructions</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">decoder</span><span class="p">((</span><span class="n">caps</span> <span class="o">*</span> <span class="n">mask</span><span class="p">[:,</span> <span class="p">:,</span> <span class="kc">None</span><span class="p">])</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="n">x</span><span class="o">.</span><span class="n">shape</span><span class="p">[</span><span class="mi">0</span><span class="p">],</span> <span class="o">-</span><span class="mi">1</span><span class="p">))</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-17'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-17'>#</a>
|
||||
</div>
|
||||
<p>重塑重建以匹配图像尺寸</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">95</span> <span class="n">reconstructions</span> <span class="o">=</span> <span class="n">reconstructions</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">28</span><span class="p">,</span> <span class="mi">28</span><span class="p">)</span>
|
||||
<span class="lineno">96</span>
|
||||
<span class="lineno">97</span> <span class="k">return</span> <span class="n">caps</span><span class="p">,</span> <span class="n">reconstructions</span><span class="p">,</span> <span class="n">pred</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-18'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-18'>#</a>
|
||||
</div>
|
||||
<p>使用 MNIST 数据和训练与验证设置的配置</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">100</span><span class="k">class</span> <span class="nc">Configs</span><span class="p">(</span><span class="n">MNISTConfigs</span><span class="p">,</span> <span class="n">SimpleTrainValidConfigs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-19'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-19'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">104</span> <span class="n">epochs</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">10</span>
|
||||
<span class="lineno">105</span> <span class="n">model</span><span class="p">:</span> <span class="n">nn</span><span class="o">.</span><span class="n">Module</span> <span class="o">=</span> <span class="s1">'capsule_network_model'</span>
|
||||
<span class="lineno">106</span> <span class="n">reconstruction_loss</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">MSELoss</span><span class="p">()</span>
|
||||
<span class="lineno">107</span> <span class="n">margin_loss</span> <span class="o">=</span> <span class="n">MarginLoss</span><span class="p">(</span><span class="n">n_labels</span><span class="o">=</span><span class="mi">10</span><span class="p">)</span>
|
||||
<span class="lineno">108</span> <span class="n">accuracy</span> <span class="o">=</span> <span class="n">AccuracyDirect</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-20'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-20'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">110</span> <span class="k">def</span> <span class="nf">init</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-21'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-21'>#</a>
|
||||
</div>
|
||||
<p>印刷损耗和屏幕精度</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">112</span> <span class="n">tracker</span><span class="o">.</span><span class="n">set_scalar</span><span class="p">(</span><span class="s1">'loss.*'</span><span class="p">,</span> <span class="kc">True</span><span class="p">)</span>
|
||||
<span class="lineno">113</span> <span class="n">tracker</span><span class="o">.</span><span class="n">set_scalar</span><span class="p">(</span><span class="s1">'accuracy.*'</span><span class="p">,</span> <span class="kc">True</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-22'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-22'>#</a>
|
||||
</div>
|
||||
<p>我们需要设置指标来计算训练和验证时期的指标</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">116</span> <span class="bp">self</span><span class="o">.</span><span class="n">state_modules</span> <span class="o">=</span> <span class="p">[</span><span class="bp">self</span><span class="o">.</span><span class="n">accuracy</span><span class="p">]</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-23'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-23'>#</a>
|
||||
</div>
|
||||
<p>这个方法被训练器调用</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">118</span> <span class="k">def</span> <span class="nf">step</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">batch</span><span class="p">:</span> <span class="n">Any</span><span class="p">,</span> <span class="n">batch_idx</span><span class="p">:</span> <span class="n">BatchIndex</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-24'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-24'>#</a>
|
||||
</div>
|
||||
<p>设置模型模式</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">123</span> <span class="bp">self</span><span class="o">.</span><span class="n">model</span><span class="o">.</span><span class="n">train</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">mode</span><span class="o">.</span><span class="n">is_train</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-25'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-25'>#</a>
|
||||
</div>
|
||||
<p>获取图像和标签并将其移动到模特的设备上</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">126</span> <span class="n">data</span><span class="p">,</span> <span class="n">target</span> <span class="o">=</span> <span class="n">batch</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">device</span><span class="p">),</span> <span class="n">batch</span><span class="p">[</span><span class="mi">1</span><span class="p">]</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">device</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-26'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-26'>#</a>
|
||||
</div>
|
||||
<p>在训练模式中增加步数</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">129</span> <span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">mode</span><span class="o">.</span><span class="n">is_train</span><span class="p">:</span>
|
||||
<span class="lineno">130</span> <span class="n">tracker</span><span class="o">.</span><span class="n">add_global_step</span><span class="p">(</span><span class="nb">len</span><span class="p">(</span><span class="n">data</span><span class="p">))</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-27'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-27'>#</a>
|
||||
</div>
|
||||
<p>是否记录激活次数</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">133</span> <span class="k">with</span> <span class="bp">self</span><span class="o">.</span><span class="n">mode</span><span class="o">.</span><span class="n">update</span><span class="p">(</span><span class="n">is_log_activations</span><span class="o">=</span><span class="n">batch_idx</span><span class="o">.</span><span class="n">is_last</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-28'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-28'>#</a>
|
||||
</div>
|
||||
<p>运行模型</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">135</span> <span class="n">caps</span><span class="p">,</span> <span class="n">reconstructions</span><span class="p">,</span> <span class="n">pred</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">model</span><span class="p">(</span><span class="n">data</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-29'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-29'>#</a>
|
||||
</div>
|
||||
<p>计算总损失</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">138</span> <span class="n">loss</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">margin_loss</span><span class="p">(</span><span class="n">caps</span><span class="p">,</span> <span class="n">target</span><span class="p">)</span> <span class="o">+</span> <span class="mf">0.0005</span> <span class="o">*</span> <span class="bp">self</span><span class="o">.</span><span class="n">reconstruction_loss</span><span class="p">(</span><span class="n">reconstructions</span><span class="p">,</span> <span class="n">data</span><span class="p">)</span>
|
||||
<span class="lineno">139</span> <span class="n">tracker</span><span class="o">.</span><span class="n">add</span><span class="p">(</span><span class="s2">"loss."</span><span class="p">,</span> <span class="n">loss</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-30'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-30'>#</a>
|
||||
</div>
|
||||
<p>呼叫准确度指标</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">142</span> <span class="bp">self</span><span class="o">.</span><span class="n">accuracy</span><span class="p">(</span><span class="n">pred</span><span class="p">,</span> <span class="n">target</span><span class="p">)</span>
|
||||
<span class="lineno">143</span>
|
||||
<span class="lineno">144</span> <span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">mode</span><span class="o">.</span><span class="n">is_train</span><span class="p">:</span>
|
||||
<span class="lineno">145</span> <span class="n">loss</span><span class="o">.</span><span class="n">backward</span><span class="p">()</span>
|
||||
<span class="lineno">146</span>
|
||||
<span class="lineno">147</span> <span class="bp">self</span><span class="o">.</span><span class="n">optimizer</span><span class="o">.</span><span class="n">step</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-31'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-31'>#</a>
|
||||
</div>
|
||||
<p>日志参数和梯度</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">149</span> <span class="k">if</span> <span class="n">batch_idx</span><span class="o">.</span><span class="n">is_last</span><span class="p">:</span>
|
||||
<span class="lineno">150</span> <span class="n">tracker</span><span class="o">.</span><span class="n">add</span><span class="p">(</span><span class="s1">'model'</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">model</span><span class="p">)</span>
|
||||
<span class="lineno">151</span> <span class="bp">self</span><span class="o">.</span><span class="n">optimizer</span><span class="o">.</span><span class="n">zero_grad</span><span class="p">()</span>
|
||||
<span class="lineno">152</span>
|
||||
<span class="lineno">153</span> <span class="n">tracker</span><span class="o">.</span><span class="n">save</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-32'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-32'>#</a>
|
||||
</div>
|
||||
<p>设置模型</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">156</span><span class="nd">@option</span><span class="p">(</span><span class="n">Configs</span><span class="o">.</span><span class="n">model</span><span class="p">)</span>
|
||||
<span class="lineno">157</span><span class="k">def</span> <span class="nf">capsule_network_model</span><span class="p">(</span><span class="n">c</span><span class="p">:</span> <span class="n">Configs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-33'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-33'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">159</span> <span class="k">return</span> <span class="n">MNISTCapsuleNetworkModel</span><span class="p">()</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">device</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-34'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-34'>#</a>
|
||||
</div>
|
||||
<p>运行实验</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">162</span><span class="k">def</span> <span class="nf">main</span><span class="p">():</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-35'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-35'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">166</span> <span class="n">experiment</span><span class="o">.</span><span class="n">create</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">'capsule_network_mnist'</span><span class="p">)</span>
|
||||
<span class="lineno">167</span> <span class="n">conf</span> <span class="o">=</span> <span class="n">Configs</span><span class="p">()</span>
|
||||
<span class="lineno">168</span> <span class="n">experiment</span><span class="o">.</span><span class="n">configs</span><span class="p">(</span><span class="n">conf</span><span class="p">,</span> <span class="p">{</span><span class="s1">'optimizer.optimizer'</span><span class="p">:</span> <span class="s1">'Adam'</span><span class="p">,</span>
|
||||
<span class="lineno">169</span> <span class="s1">'optimizer.learning_rate'</span><span class="p">:</span> <span class="mf">1e-3</span><span class="p">})</span>
|
||||
<span class="lineno">170</span>
|
||||
<span class="lineno">171</span> <span class="n">experiment</span><span class="o">.</span><span class="n">add_pytorch_models</span><span class="p">({</span><span class="s1">'model'</span><span class="p">:</span> <span class="n">conf</span><span class="o">.</span><span class="n">model</span><span class="p">})</span>
|
||||
<span class="lineno">172</span>
|
||||
<span class="lineno">173</span> <span class="k">with</span> <span class="n">experiment</span><span class="o">.</span><span class="n">start</span><span class="p">():</span>
|
||||
<span class="lineno">174</span> <span class="n">conf</span><span class="o">.</span><span class="n">run</span><span class="p">()</span>
|
||||
<span class="lineno">175</span>
|
||||
<span class="lineno">176</span>
|
||||
<span class="lineno">177</span><span class="k">if</span> <span class="vm">__name__</span> <span class="o">==</span> <span class="s1">'__main__'</span><span class="p">:</span>
|
||||
<span class="lineno">178</span> <span class="n">main</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
@@ -0,0 +1,132 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content=""/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="胶囊网络"/>
|
||||
<meta name="twitter:description" content=""/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/capsule_networks/readme.html"/>
|
||||
<meta property="og:title" content="胶囊网络"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="胶囊网络"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="胶囊网络"/>
|
||||
<meta property="og:description" content=""/>
|
||||
|
||||
<title>胶囊网络</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/capsule_networks/readme.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="index.html">capsule_networks</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/capsule_networks/readme.md" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1><a href="https://nn.labml.ai/capsule_networks/index.html">胶囊网络</a></h1>
|
||||
<p>这是<a href="https://arxiv.org/abs/1710.09829">胶囊间动态路由</a>的 <a href="https://pytorch.org">PyTorch</a> 实现/教程。</p>
|
||||
<p>Capsule 网络是一种神经网络架构，它以胶囊的形式嵌入特征，并通过投票机制将它们路由到下一层胶囊。</p>
|
||||
<p>与其他模型实现不同，我们提供了一个示例，因为仅使用模块很难理解某些概念。<a href="mnist.html">这是使用胶囊对 MNIST 数据集进行分类的模型的带注释的代码</a></p>
|
||||
<p>该文件包含了 Capsule Networks 核心模块的实现。</p>
|
||||
<p>我用 <a href="https://github.com/jindongwang/Pytorch-CapsuleNet">jindongwang/pytorch-CapsuleNet</a> 来澄清我对这篇论文的一些困惑。</p>
|
||||
<p>这是一本在 MNIST 数据集上训练 Capsule 网络的笔记本。</p>
|
||||
<p><a href="https://colab.research.google.com/github/labmlai/annotated_deep_learning_paper_implementations/blob/master/labml_nn/capsule_networks/mnist.ipynb"><img alt="Open In Colab" src="https://colab.research.google.com/assets/colab-badge.svg"></a></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
@@ -0,0 +1,205 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content=""/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="analytics.py"/>
|
||||
<meta name="twitter:description" content=""/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/cfr/analytics.html"/>
|
||||
<meta property="og:title" content="analytics.py"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="analytics.py"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="analytics.py"/>
|
||||
<meta property="og:description" content=""/>
|
||||
|
||||
<title>analytics.py</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/cfr/analytics.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="index.html">cfr</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/cfr/analytics.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">1</span><span></span><span class="kn">from</span> <span class="nn">typing</span> <span class="kn">import</span> <span class="n">List</span>
|
||||
<span class="lineno">2</span>
|
||||
<span class="lineno">3</span><span class="kn">import</span> <span class="nn">altair</span> <span class="k">as</span> <span class="nn">alt</span>
|
||||
<span class="lineno">4</span><span class="kn">import</span> <span class="nn">numpy</span> <span class="k">as</span> <span class="nn">np</span>
|
||||
<span class="lineno">5</span>
|
||||
<span class="lineno">6</span><span class="kn">from</span> <span class="nn">labml</span> <span class="kn">import</span> <span class="n">analytics</span>
|
||||
<span class="lineno">7</span><span class="kn">from</span> <span class="nn">labml.analytics</span> <span class="kn">import</span> <span class="n">IndicatorCollection</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-1'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-1'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">10</span><span class="k">def</span> <span class="nf">calculate_percentages</span><span class="p">(</span><span class="n">means</span><span class="p">:</span> <span class="n">List</span><span class="p">[</span><span class="n">np</span><span class="o">.</span><span class="n">ndarray</span><span class="p">],</span> <span class="n">names</span><span class="p">:</span> <span class="n">List</span><span class="p">[</span><span class="n">List</span><span class="p">[</span><span class="nb">str</span><span class="p">]]):</span>
|
||||
<span class="lineno">11</span> <span class="n">normalized</span> <span class="o">=</span> <span class="p">[]</span>
|
||||
<span class="lineno">12</span>
|
||||
<span class="lineno">13</span> <span class="k">for</span> <span class="n">i</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="nb">len</span><span class="p">(</span><span class="n">means</span><span class="p">)):</span>
|
||||
<span class="lineno">14</span> <span class="n">total</span> <span class="o">=</span> <span class="n">np</span><span class="o">.</span><span class="n">zeros_like</span><span class="p">(</span><span class="n">means</span><span class="p">[</span><span class="n">i</span><span class="p">])</span>
|
||||
<span class="lineno">15</span> <span class="k">for</span> <span class="n">j</span><span class="p">,</span> <span class="n">n</span> <span class="ow">in</span> <span class="nb">enumerate</span><span class="p">(</span><span class="n">names</span><span class="p">):</span>
|
||||
<span class="lineno">16</span> <span class="k">if</span> <span class="n">n</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">][:</span><span class="o">-</span><span class="mi">1</span><span class="p">]</span> <span class="o">==</span> <span class="n">names</span><span class="p">[</span><span class="n">i</span><span class="p">][</span><span class="o">-</span><span class="mi">1</span><span class="p">][:</span><span class="o">-</span><span class="mi">1</span><span class="p">]:</span>
|
||||
<span class="lineno">17</span> <span class="n">total</span> <span class="o">+=</span> <span class="n">means</span><span class="p">[</span><span class="n">j</span><span class="p">]</span>
|
||||
<span class="lineno">18</span> <span class="n">normalized</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">means</span><span class="p">[</span><span class="n">i</span><span class="p">]</span> <span class="o">/</span> <span class="p">(</span><span class="n">total</span> <span class="o">+</span> <span class="n">np</span><span class="o">.</span><span class="n">finfo</span><span class="p">(</span><span class="nb">float</span><span class="p">)</span><span class="o">.</span><span class="n">eps</span><span class="p">))</span>
|
||||
<span class="lineno">19</span>
|
||||
<span class="lineno">20</span> <span class="k">return</span> <span class="n">normalized</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-2'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-2'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">23</span><span class="k">def</span> <span class="nf">plot_infosets</span><span class="p">(</span><span class="n">indicators</span><span class="p">:</span> <span class="n">IndicatorCollection</span><span class="p">,</span> <span class="o">*</span><span class="p">,</span>
|
||||
<span class="lineno">24</span> <span class="n">is_normalize</span><span class="p">:</span> <span class="nb">bool</span> <span class="o">=</span> <span class="kc">True</span><span class="p">,</span>
|
||||
<span class="lineno">25</span> <span class="n">width</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">600</span><span class="p">,</span>
|
||||
<span class="lineno">26</span> <span class="n">height</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">300</span><span class="p">):</span>
|
||||
<span class="lineno">27</span> <span class="n">data</span><span class="p">,</span> <span class="n">names</span> <span class="o">=</span> <span class="n">analytics</span><span class="o">.</span><span class="n">indicator_data</span><span class="p">(</span><span class="n">indicators</span><span class="p">)</span>
|
||||
<span class="lineno">28</span> <span class="n">step</span> <span class="o">=</span> <span class="p">[</span><span class="n">d</span><span class="p">[:,</span> <span class="mi">0</span><span class="p">]</span> <span class="k">for</span> <span class="n">d</span> <span class="ow">in</span> <span class="n">data</span><span class="p">]</span>
|
||||
<span class="lineno">29</span> <span class="n">means</span> <span class="o">=</span> <span class="p">[</span><span class="n">d</span><span class="p">[:,</span> <span class="mi">5</span><span class="p">]</span> <span class="k">for</span> <span class="n">d</span> <span class="ow">in</span> <span class="n">data</span><span class="p">]</span>
|
||||
<span class="lineno">30</span>
|
||||
<span class="lineno">31</span> <span class="k">if</span> <span class="n">is_normalize</span><span class="p">:</span>
|
||||
<span class="lineno">32</span> <span class="n">normalized</span> <span class="o">=</span> <span class="n">calculate_percentages</span><span class="p">(</span><span class="n">means</span><span class="p">,</span> <span class="n">names</span><span class="p">)</span>
|
||||
<span class="lineno">33</span> <span class="k">else</span><span class="p">:</span>
|
||||
<span class="lineno">34</span> <span class="n">normalized</span> <span class="o">=</span> <span class="n">means</span>
|
||||
<span class="lineno">35</span>
|
||||
<span class="lineno">36</span> <span class="n">common</span> <span class="o">=</span> <span class="n">names</span><span class="p">[</span><span class="mi">0</span><span class="p">][</span><span class="o">-</span><span class="mi">1</span><span class="p">]</span>
|
||||
<span class="lineno">37</span> <span class="k">for</span> <span class="n">i</span><span class="p">,</span> <span class="n">n</span> <span class="ow">in</span> <span class="nb">enumerate</span><span class="p">(</span><span class="n">names</span><span class="p">):</span>
|
||||
<span class="lineno">38</span> <span class="n">n</span> <span class="o">=</span> <span class="n">n</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">]</span>
|
||||
<span class="lineno">39</span> <span class="k">if</span> <span class="nb">len</span><span class="p">(</span><span class="n">n</span><span class="p">)</span> <span class="o"><</span> <span class="nb">len</span><span class="p">(</span><span class="n">common</span><span class="p">):</span>
|
||||
<span class="lineno">40</span> <span class="n">common</span> <span class="o">=</span> <span class="n">common</span><span class="p">[:</span><span class="nb">len</span><span class="p">(</span><span class="n">n</span><span class="p">)]</span>
|
||||
<span class="lineno">41</span> <span class="k">for</span> <span class="n">j</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="nb">len</span><span class="p">(</span><span class="n">common</span><span class="p">)):</span>
|
||||
<span class="lineno">42</span> <span class="k">if</span> <span class="n">common</span><span class="p">[</span><span class="n">j</span><span class="p">]</span> <span class="o">!=</span> <span class="n">n</span><span class="p">[</span><span class="n">j</span><span class="p">]:</span>
|
||||
<span class="lineno">43</span> <span class="n">common</span> <span class="o">=</span> <span class="n">common</span><span class="p">[:</span><span class="n">j</span><span class="p">]</span>
|
||||
<span class="lineno">44</span> <span class="k">break</span>
|
||||
<span class="lineno">45</span>
|
||||
<span class="lineno">46</span> <span class="n">table</span> <span class="o">=</span> <span class="p">[]</span>
|
||||
<span class="lineno">47</span> <span class="k">for</span> <span class="n">i</span><span class="p">,</span> <span class="n">n</span> <span class="ow">in</span> <span class="nb">enumerate</span><span class="p">(</span><span class="n">names</span><span class="p">):</span>
|
||||
<span class="lineno">48</span> <span class="k">for</span> <span class="n">j</span><span class="p">,</span> <span class="n">v</span> <span class="ow">in</span> <span class="nb">zip</span><span class="p">(</span><span class="n">step</span><span class="p">[</span><span class="n">i</span><span class="p">],</span> <span class="n">normalized</span><span class="p">[</span><span class="n">i</span><span class="p">]):</span>
|
||||
<span class="lineno">49</span> <span class="n">table</span><span class="o">.</span><span class="n">append</span><span class="p">({</span>
|
||||
<span class="lineno">50</span> <span class="s1">'series'</span><span class="p">:</span> <span class="n">n</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">][</span><span class="nb">len</span><span class="p">(</span><span class="n">common</span><span class="p">):],</span>
|
||||
<span class="lineno">51</span> <span class="s1">'step'</span><span class="p">:</span> <span class="n">j</span><span class="p">,</span>
|
||||
<span class="lineno">52</span> <span class="s1">'value'</span><span class="p">:</span> <span class="n">v</span>
|
||||
<span class="lineno">53</span> <span class="p">})</span>
|
||||
<span class="lineno">54</span>
|
||||
<span class="lineno">55</span> <span class="n">table</span> <span class="o">=</span> <span class="n">alt</span><span class="o">.</span><span class="n">Data</span><span class="p">(</span><span class="n">values</span><span class="o">=</span><span class="n">table</span><span class="p">)</span>
|
||||
<span class="lineno">56</span>
|
||||
<span class="lineno">57</span> <span class="n">selection</span> <span class="o">=</span> <span class="n">alt</span><span class="o">.</span><span class="n">selection_multi</span><span class="p">(</span><span class="n">fields</span><span class="o">=</span><span class="p">[</span><span class="s1">'series'</span><span class="p">],</span> <span class="n">bind</span><span class="o">=</span><span class="s1">'legend'</span><span class="p">)</span>
|
||||
<span class="lineno">58</span>
|
||||
<span class="lineno">59</span> <span class="k">return</span> <span class="n">alt</span><span class="o">.</span><span class="n">Chart</span><span class="p">(</span><span class="n">table</span><span class="p">)</span><span class="o">.</span><span class="n">mark_line</span><span class="p">()</span><span class="o">.</span><span class="n">encode</span><span class="p">(</span>
|
||||
<span class="lineno">60</span> <span class="n">alt</span><span class="o">.</span><span class="n">X</span><span class="p">(</span><span class="s1">'step:Q'</span><span class="p">),</span>
|
||||
<span class="lineno">61</span> <span class="n">alt</span><span class="o">.</span><span class="n">Y</span><span class="p">(</span><span class="s1">'value:Q'</span><span class="p">),</span>
|
||||
<span class="lineno">62</span> <span class="n">alt</span><span class="o">.</span><span class="n">Color</span><span class="p">(</span><span class="s1">'series:N'</span><span class="p">,</span> <span class="n">scale</span><span class="o">=</span><span class="n">alt</span><span class="o">.</span><span class="n">Scale</span><span class="p">(</span><span class="n">scheme</span><span class="o">=</span><span class="s1">'tableau20'</span><span class="p">)),</span>
|
||||
<span class="lineno">63</span> <span class="n">opacity</span><span class="o">=</span><span class="n">alt</span><span class="o">.</span><span class="n">condition</span><span class="p">(</span><span class="n">selection</span><span class="p">,</span> <span class="n">alt</span><span class="o">.</span><span class="n">value</span><span class="p">(</span><span class="mi">1</span><span class="p">),</span> <span class="n">alt</span><span class="o">.</span><span class="n">value</span><span class="p">(</span><span class="mf">0.0001</span><span class="p">))</span>
|
||||
<span class="lineno">64</span> <span class="p">)</span><span class="o">.</span><span class="n">add_selection</span><span class="p">(</span>
|
||||
<span class="lineno">65</span> <span class="n">selection</span>
|
||||
<span class="lineno">66</span> <span class="p">)</span><span class="o">.</span><span class="n">properties</span><span class="p">(</span><span class="n">width</span><span class="o">=</span><span class="n">width</span><span class="p">,</span> <span class="n">height</span><span class="o">=</span><span class="n">height</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
File diff suppressed because one or more lines are too long
@@ -0,0 +1,186 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content=""/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="infoset_saver.py"/>
|
||||
<meta name="twitter:description" content=""/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/cfr/infoset_saver.html"/>
|
||||
<meta property="og:title" content="infoset_saver.py"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="infoset_saver.py"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="infoset_saver.py"/>
|
||||
<meta property="og:description" content=""/>
|
||||
|
||||
<title>infoset_saver.py</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/cfr/infoset_saver.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="index.html">cfr</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/cfr/infoset_saver.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">1</span><span></span><span class="kn">import</span> <span class="nn">json</span>
|
||||
<span class="lineno">2</span><span class="kn">import</span> <span class="nn">pathlib</span>
|
||||
<span class="lineno">3</span><span class="kn">from</span> <span class="nn">typing</span> <span class="kn">import</span> <span class="n">Dict</span>
|
||||
<span class="lineno">4</span>
|
||||
<span class="lineno">5</span><span class="kn">from</span> <span class="nn">labml</span> <span class="kn">import</span> <span class="n">experiment</span>
|
||||
<span class="lineno">6</span><span class="kn">from</span> <span class="nn">labml_nn.cfr</span> <span class="kn">import</span> <span class="n">InfoSet</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-1'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-1'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">9</span><span class="k">class</span> <span class="nc">InfoSetSaver</span><span class="p">(</span><span class="n">experiment</span><span class="o">.</span><span class="n">ModelSaver</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-2'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-2'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">10</span> <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">infosets</span><span class="p">:</span> <span class="n">Dict</span><span class="p">[</span><span class="nb">str</span><span class="p">,</span> <span class="n">InfoSet</span><span class="p">]):</span>
|
||||
<span class="lineno">11</span> <span class="bp">self</span><span class="o">.</span><span class="n">infosets</span> <span class="o">=</span> <span class="n">infosets</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-3'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-3'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">13</span> <span class="k">def</span> <span class="nf">save</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">checkpoint_path</span><span class="p">:</span> <span class="n">pathlib</span><span class="o">.</span><span class="n">Path</span><span class="p">)</span> <span class="o">-></span> <span class="nb">any</span><span class="p">:</span>
|
||||
<span class="lineno">14</span> <span class="n">data</span> <span class="o">=</span> <span class="p">{</span><span class="n">key</span><span class="p">:</span> <span class="n">infoset</span><span class="o">.</span><span class="n">to_dict</span><span class="p">()</span> <span class="k">for</span> <span class="n">key</span><span class="p">,</span> <span class="n">infoset</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">infosets</span><span class="o">.</span><span class="n">items</span><span class="p">()}</span>
|
||||
<span class="lineno">15</span> <span class="n">file_name</span> <span class="o">=</span> <span class="sa">f</span><span class="s2">"infosets.json"</span>
|
||||
<span class="lineno">16</span>
|
||||
<span class="lineno">17</span> <span class="k">with</span> <span class="nb">open</span><span class="p">(</span><span class="nb">str</span><span class="p">(</span><span class="n">checkpoint_path</span> <span class="o">/</span> <span class="n">file_name</span><span class="p">),</span> <span class="s1">'w'</span><span class="p">)</span> <span class="k">as</span> <span class="n">f</span><span class="p">:</span>
|
||||
<span class="lineno">18</span> <span class="n">f</span><span class="o">.</span><span class="n">write</span><span class="p">(</span><span class="n">json</span><span class="o">.</span><span class="n">dumps</span><span class="p">(</span><span class="n">data</span><span class="p">))</span>
|
||||
<span class="lineno">19</span>
|
||||
<span class="lineno">20</span> <span class="k">return</span> <span class="n">file_name</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-4'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-4'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">22</span> <span class="k">def</span> <span class="nf">load</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">checkpoint_path</span><span class="p">:</span> <span class="n">pathlib</span><span class="o">.</span><span class="n">Path</span><span class="p">,</span> <span class="n">file_name</span><span class="p">:</span> <span class="nb">str</span><span class="p">):</span>
|
||||
<span class="lineno">23</span> <span class="k">with</span> <span class="nb">open</span><span class="p">(</span><span class="nb">str</span><span class="p">(</span><span class="n">checkpoint_path</span> <span class="o">/</span> <span class="n">file_name</span><span class="p">),</span> <span class="s1">'w'</span><span class="p">)</span> <span class="k">as</span> <span class="n">f</span><span class="p">:</span>
|
||||
<span class="lineno">24</span> <span class="n">data</span> <span class="o">=</span> <span class="n">json</span><span class="o">.</span><span class="n">loads</span><span class="p">(</span><span class="n">f</span><span class="o">.</span><span class="n">read</span><span class="p">())</span>
|
||||
<span class="lineno">25</span>
|
||||
<span class="lineno">26</span> <span class="k">for</span> <span class="n">key</span><span class="p">,</span> <span class="n">d</span> <span class="ow">in</span> <span class="n">data</span><span class="o">.</span><span class="n">items</span><span class="p">():</span>
|
||||
<span class="lineno">27</span> <span class="bp">self</span><span class="o">.</span><span class="n">infosets</span><span class="p">[</span><span class="n">key</span><span class="p">]</span> <span class="o">=</span> <span class="n">InfoSet</span><span class="o">.</span><span class="n">from_dict</span><span class="p">(</span><span class="n">d</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
@@ -0,0 +1,874 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content="这是 Kuhn Poker 上的 CFR 带注释的实现/教程"/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="库恩扑克上的 CFR"/>
|
||||
<meta name="twitter:description" content="这是 Kuhn Poker 上的 CFR 带注释的实现/教程"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/cfr/kuhn/index.html"/>
|
||||
<meta property="og:title" content="库恩扑克上的 CFR"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="库恩扑克上的 CFR"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="库恩扑克上的 CFR"/>
|
||||
<meta property="og:description" content="这是 Kuhn Poker 上的 CFR 带注释的实现/教程"/>
|
||||
|
||||
<title>库恩扑克上的 CFR</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/cfr/kuhn/index.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="../index.html">cfr</a>
|
||||
<a class="parent" href="index.html">kuhn</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/cfr/kuhn/__init__.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1>Kuhn Poker 上的<a href="../index.html">反事实后悔最小化 (CFR)</a></h1>
|
||||
<p>这将《<a href="../index.html">反事实后悔最小化》（CFR）</a>应用于库恩扑克。</p>
|
||||
<p><a href="https://en.wikipedia.org/wiki/Kuhn_poker">Kuhn Poker</a> 是一款双人三张牌的下注游戏。玩家从 A、K 和 Queen 中各获得一张牌（无花色）。包中只有三张牌，所以遗漏了一张牌。Ace 击败 King 和 Queen，King 击败 Queen ——就像普通卡牌排名一样。</p>
|
||||
<p>两个玩家都押注<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqe" style=""><span class="mord" style="">1</span></span></span></span></span></span>筹码（盲注<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqe" style=""><span class="mord" style="">1</span></span></span></span></span></span>筹码）。看完牌后，第一个玩家可以传球或下注<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqe" style=""><span class="mord" style="">1</span></span></span></span></span></span>筹码。如果第一个玩家通过，则拥有较高牌的玩家赢得底池。如果第一个玩家下注，第二个玩家可以下注（即跟注）<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqe" style=""><span class="mord" style="">1</span></span></span></span></span></span>筹码或过关（即盖牌）。如果第二个玩家下注，而拥有较高牌的玩家赢得底池。如果第二个玩家通过（即弃牌），则第一个玩家获得底池。这个游戏是反复玩的，一个好的策略将针对长期效用（或获利）进行优化。</p>
|
||||
<p>以下是一些示例游戏：</p>
|
||||
<ul><li><code class="highlight"><span></span><span class="n">KAp</span></code>
|
||||
-玩家 1 获得 K 玩家 2 获得 A. 玩家 1 传球。玩家 2 没有下注机会，玩家 2 赢得<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqf" style=""><span class="mord" style="">2</span></span></span></span></span></span>筹码底池。</li>
|
||||
<li><code class="highlight"><span></span><span class="n">QKbp</span></code>
|
||||
-玩家 1 得到 Q。玩家 2 得到 K。玩家 1 下注一个筹码。玩家2传球（弃牌）。玩家 1 拿到底池<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqg" style=""><span class="mord" style="">4</span></span></span></span></span></span>是因为玩家 2 盖牌。</li>
|
||||
<li><code class="highlight"><span></span><span class="n">QAbb</span></code>
|
||||
-玩家 1 得到 Q。玩家 2 得到 A. 玩家 1 下注一个筹码。玩家2也下注（跟注）。玩家 2 赢得底池<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqg" style=""><span class="mord" style="">4</span></span></span></span></span></span>。</li></ul>
|
||||
<p>他<a href="../index.html"><code class="highlight"><span></span><span class="fm">__init__</span><span class="o">.</span><span class="n">py</span></code>
|
||||
</a>用Kuhn Poker的具体内容扩展了中定义的<code class="highlight"><span></span><span class="n">InfoSet</span></code>
|
||||
<code class="highlight"><span></span><span class="n">History</span></code>
|
||||
类别和等级。</p>
|
||||
<p><a href="https://colab.research.google.com/github/labmlai/annotated_deep_learning_paper_implementations/blob/master/labml_nn/cfr/kuhn/experiment.ipynb"><img alt="Open In Colab" src="https://colab.research.google.com/assets/colab-badge.svg"></a></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">37</span><span></span><span class="kn">from</span> <span class="nn">typing</span> <span class="kn">import</span> <span class="n">List</span><span class="p">,</span> <span class="n">cast</span><span class="p">,</span> <span class="n">Dict</span>
|
||||
<span class="lineno">38</span>
|
||||
<span class="lineno">39</span><span class="kn">import</span> <span class="nn">numpy</span> <span class="k">as</span> <span class="nn">np</span>
|
||||
<span class="lineno">40</span>
|
||||
<span class="lineno">41</span><span class="kn">from</span> <span class="nn">labml</span> <span class="kn">import</span> <span class="n">experiment</span>
|
||||
<span class="lineno">42</span><span class="kn">from</span> <span class="nn">labml.configs</span> <span class="kn">import</span> <span class="n">option</span>
|
||||
<span class="lineno">43</span><span class="kn">from</span> <span class="nn">labml_nn.cfr</span> <span class="kn">import</span> <span class="n">History</span> <span class="k">as</span> <span class="n">_History</span><span class="p">,</span> <span class="n">InfoSet</span> <span class="k">as</span> <span class="n">_InfoSet</span><span class="p">,</span> <span class="n">Action</span><span class="p">,</span> <span class="n">Player</span><span class="p">,</span> <span class="n">CFRConfigs</span>
|
||||
<span class="lineno">44</span><span class="kn">from</span> <span class="nn">labml_nn.cfr.infoset_saver</span> <span class="kn">import</span> <span class="n">InfoSetSaver</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-1'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-1'>#</a>
|
||||
</div>
|
||||
<p>库恩扑克的动作是 pass (<code class="highlight"><span></span><span class="n">p</span></code>
|
||||
) 或 bet (<code class="highlight"><span></span><span class="n">b</span></code>
|
||||
)</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">47</span><span class="n">ACTIONS</span> <span class="o">=</span> <span class="n">cast</span><span class="p">(</span><span class="n">List</span><span class="p">[</span><span class="n">Action</span><span class="p">],</span> <span class="p">[</span><span class="s1">'p'</span><span class="p">,</span> <span class="s1">'b'</span><span class="p">])</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-2'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-2'>#</a>
|
||||
</div>
|
||||
<p>游戏中的三张牌分别是 Ace、King 和 Queen</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">49</span><span class="n">CHANCES</span> <span class="o">=</span> <span class="n">cast</span><span class="p">(</span><span class="n">List</span><span class="p">[</span><span class="n">Action</span><span class="p">],</span> <span class="p">[</span><span class="s1">'A'</span><span class="p">,</span> <span class="s1">'K'</span><span class="p">,</span> <span class="s1">'Q'</span><span class="p">])</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-3'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-3'>#</a>
|
||||
</div>
|
||||
<p>有两个玩家</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">51</span><span class="n">PLAYERS</span> <span class="o">=</span> <span class="n">cast</span><span class="p">(</span><span class="n">List</span><span class="p">[</span><span class="n">Player</span><span class="p">],</span> <span class="p">[</span><span class="mi">0</span><span class="p">,</span> <span class="mi">1</span><span class="p">])</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-4'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-4'>#</a>
|
||||
</div>
|
||||
<h2><a href="../index.html#InfoSet">信息集</a></h2>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">54</span><span class="k">class</span> <span class="nc">InfoSet</span><span class="p">(</span><span class="n">_InfoSet</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-5'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-5'>#</a>
|
||||
</div>
|
||||
<p>不支持保存/加载</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">59</span> <span class="nd">@staticmethod</span>
|
||||
<span class="lineno">60</span> <span class="k">def</span> <span class="nf">from_dict</span><span class="p">(</span><span class="n">data</span><span class="p">:</span> <span class="n">Dict</span><span class="p">[</span><span class="nb">str</span><span class="p">,</span> <span class="nb">any</span><span class="p">])</span> <span class="o">-></span> <span class="s1">'InfoSet'</span><span class="p">:</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-6'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-6'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">62</span> <span class="k">pass</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-7'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-7'>#</a>
|
||||
</div>
|
||||
<p>返回操作列表。终端状态由<code class="highlight"><span></span><span class="n">History</span></code>
|
||||
类处理。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">64</span> <span class="k">def</span> <span class="nf">actions</span><span class="p">(</span><span class="bp">self</span><span class="p">)</span> <span class="o">-></span> <span class="n">List</span><span class="p">[</span><span class="n">Action</span><span class="p">]:</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-8'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-8'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">68</span> <span class="k">return</span> <span class="n">ACTIONS</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-9'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-9'>#</a>
|
||||
</div>
|
||||
<p>人类可读的字符串表示——它给出了下注概率</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">70</span> <span class="k">def</span> <span class="fm">__repr__</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-10'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-10'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">74</span> <span class="n">total</span> <span class="o">=</span> <span class="nb">sum</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">cumulative_strategy</span><span class="o">.</span><span class="n">values</span><span class="p">())</span>
|
||||
<span class="lineno">75</span> <span class="n">total</span> <span class="o">=</span> <span class="nb">max</span><span class="p">(</span><span class="n">total</span><span class="p">,</span> <span class="mf">1e-6</span><span class="p">)</span>
|
||||
<span class="lineno">76</span> <span class="n">bet</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">cumulative_strategy</span><span class="p">[</span><span class="n">cast</span><span class="p">(</span><span class="n">Action</span><span class="p">,</span> <span class="s1">'b'</span><span class="p">)]</span> <span class="o">/</span> <span class="n">total</span>
|
||||
<span class="lineno">77</span> <span class="k">return</span> <span class="sa">f</span><span class="s1">'</span><span class="si">{</span><span class="n">bet</span><span class="w"> </span><span class="o">*</span><span class="w"> </span><span class="mi">100</span><span class="si">:</span><span class="s1"> .1f</span><span class="si">}</span><span class="s1">%'</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-11'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-11'>#</a>
|
||||
</div>
|
||||
<h2><a href="../index.html#History">历史</a></h2>
|
||||
<p>这定义了游戏何时结束，计算效用和抽样机会事件（发牌）。</p>
|
||||
<p>历史记录存储在字符串中：</p>
|
||||
<ul><li>前两个角色是发给玩家 1 和玩家 2 的牌</li>
|
||||
<li>第三个角色是第一个玩家的动作</li>
|
||||
<li>第四个角色是第二个玩家的动作</li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">80</span><span class="k">class</span> <span class="nc">History</span><span class="p">(</span><span class="n">_History</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-12'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-12'>#</a>
|
||||
</div>
|
||||
<p>历史</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">94</span> <span class="n">history</span><span class="p">:</span> <span class="nb">str</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-13'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-13'>#</a>
|
||||
</div>
|
||||
<p>使用给定的历史字符串进行初始化</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">96</span> <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">history</span><span class="p">:</span> <span class="nb">str</span> <span class="o">=</span> <span class="s1">''</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-14'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-14'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">100</span> <span class="bp">self</span><span class="o">.</span><span class="n">history</span> <span class="o">=</span> <span class="n">history</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-15'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-15'>#</a>
|
||||
</div>
|
||||
<p>历史记录是否终止（游戏结束）。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">102</span> <span class="k">def</span> <span class="nf">is_terminal</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-16'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-16'>#</a>
|
||||
</div>
|
||||
<p>玩家尚未采取行动</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">107</span> <span class="k">if</span> <span class="nb">len</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">history</span><span class="p">)</span> <span class="o"><=</span> <span class="mi">2</span><span class="p">:</span>
|
||||
<span class="lineno">108</span> <span class="k">return</span> <span class="kc">False</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-17'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-17'>#</a>
|
||||
</div>
|
||||
<p>最后一个玩过的玩家通过（游戏结束）</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">110</span> <span class="k">elif</span> <span class="bp">self</span><span class="o">.</span><span class="n">history</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">]</span> <span class="o">==</span> <span class="s1">'p'</span><span class="p">:</span>
|
||||
<span class="lineno">111</span> <span class="k">return</span> <span class="kc">True</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-18'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-18'>#</a>
|
||||
</div>
|
||||
<p>两位玩家都叫（下注）（游戏结束）</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">113</span> <span class="k">elif</span> <span class="bp">self</span><span class="o">.</span><span class="n">history</span><span class="p">[</span><span class="o">-</span><span class="mi">2</span><span class="p">:]</span> <span class="o">==</span> <span class="s1">'bb'</span><span class="p">:</span>
|
||||
<span class="lineno">114</span> <span class="k">return</span> <span class="kc">True</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-19'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-19'>#</a>
|
||||
</div>
|
||||
<p>任何其他组合</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">116</span> <span class="k">else</span><span class="p">:</span>
|
||||
<span class="lineno">117</span> <span class="k">return</span> <span class="kc">False</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-20'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-20'>#</a>
|
||||
</div>
|
||||
<p>计算玩家的终端实用程序<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqe" style=""><span class="mord" style="">1</span></span></span></span></span></span>，<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:1em;vertical-align:-0.25em;"></span><span class="mord coloredeq eqb" style=""><span class="mord" style=""><span class="mord mathnormal" style="">u</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.30110799999999993em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style=""><span class="mord mtight coloredeq eqe" style="">1</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span><span class="mopen" style="">(</span><span class="mord mathnormal" style="margin-right:0.04398em">z</span><span class="mclose" style="">)</span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">119</span> <span class="k">def</span> <span class="nf">_terminal_utility_p1</span><span class="p">(</span><span class="bp">self</span><span class="p">)</span> <span class="o">-></span> <span class="nb">float</span><span class="p">:</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-21'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-21'>#</a>
|
||||
</div>
|
||||
<p><span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.72777em;vertical-align:-0.08333em;"></span><span class="mord">+</span><span class="mord coloredeq eqe" style=""><span class="mord" style="">1</span></span></span></span></span></span>如果玩家 1 有更好的牌<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.72777em;vertical-align:-0.08333em;"></span><span class="mord">−</span><span class="mord coloredeq eqe" style=""><span class="mord" style="">1</span></span></span></span></span></span>或者其他</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">124</span> <span class="n">winner</span> <span class="o">=</span> <span class="o">-</span><span class="mi">1</span> <span class="o">+</span> <span class="mi">2</span> <span class="o">*</span> <span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">history</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span> <span class="o"><</span> <span class="bp">self</span><span class="o">.</span><span class="n">history</span><span class="p">[</span><span class="mi">1</span><span class="p">])</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-22'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-22'>#</a>
|
||||
</div>
|
||||
<p>第二个玩家通过了</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">127</span> <span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">history</span><span class="p">[</span><span class="o">-</span><span class="mi">2</span><span class="p">:]</span> <span class="o">==</span> <span class="s1">'bp'</span><span class="p">:</span>
|
||||
<span class="lineno">128</span> <span class="k">return</span> <span class="mi">1</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-23'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-23'>#</a>
|
||||
</div>
|
||||
<p>两位玩家都被召唤，持有更好牌的玩家将赢得<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqf" style=""><span class="mord" style="">2</span></span></span></span></span></span>筹码</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">130</span> <span class="k">elif</span> <span class="bp">self</span><span class="o">.</span><span class="n">history</span><span class="p">[</span><span class="o">-</span><span class="mi">2</span><span class="p">:]</span> <span class="o">==</span> <span class="s1">'bb'</span><span class="p">:</span>
|
||||
<span class="lineno">131</span> <span class="k">return</span> <span class="n">winner</span> <span class="o">*</span> <span class="mi">2</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-24'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-24'>#</a>
|
||||
</div>
|
||||
<p>第一个玩家通过，持有更好牌的玩家将赢得<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord coloredeq eqe" style=""><span class="mord" style="">1</span></span></span></span></span></span>筹码</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">133</span> <span class="k">elif</span> <span class="bp">self</span><span class="o">.</span><span class="n">history</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">]</span> <span class="o">==</span> <span class="s1">'p'</span><span class="p">:</span>
|
||||
<span class="lineno">134</span> <span class="k">return</span> <span class="n">winner</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-25'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-25'>#</a>
|
||||
</div>
|
||||
<p>历史不是终结的</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">136</span> <span class="k">else</span><span class="p">:</span>
|
||||
<span class="lineno">137</span> <span class="k">raise</span> <span class="ne">RuntimeError</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-26'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-26'>#</a>
|
||||
</div>
|
||||
<p>获取玩家的终端实用程序<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.65952em;vertical-align:0em;"></span><span class="mord coloredeq eqh" style=""><span class="mord mathnormal" style="">i</span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">139</span> <span class="k">def</span> <span class="nf">terminal_utility</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">i</span><span class="p">:</span> <span class="n">Player</span><span class="p">)</span> <span class="o">-></span> <span class="nb">float</span><span class="p">:</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-27'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-27'>#</a>
|
||||
</div>
|
||||
<p>如果<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.65952em;vertical-align:0em;"></span><span class="mord coloredeq eqh" style=""><span class="mord mathnormal" style="">i</span></span></span></span></span></span>是玩家 1</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">144</span> <span class="k">if</span> <span class="n">i</span> <span class="o">==</span> <span class="n">PLAYERS</span><span class="p">[</span><span class="mi">0</span><span class="p">]:</span>
|
||||
<span class="lineno">145</span> <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">_terminal_utility_p1</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-28'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-28'>#</a>
|
||||
</div>
|
||||
<p>否则，<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:1em;vertical-align:-0.25em;"></span><span class="mord"><span class="mord mathnormal">u</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.30110799999999993em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight"><span class="mord mtight coloredeq eqf" style=""><span class="mord mtight" style="">2</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span><span class="mopen">(</span><span class="mord mathnormal" style="margin-right:0.04398em;">z</span><span class="mclose">)</span><span class="mspace" style="margin-right:0.2777777777777778em;"></span><span class="mrel">=</span><span class="mspace" style="margin-right:0.2777777777777778em;"></span></span><span class="base"><span class="strut" style="height:1em;vertical-align:-0.25em;"></span><span class="mord">−</span><span class="mord coloredeq eqb" style=""><span class="mord" style=""><span class="mord mathnormal" style="">u</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.30110799999999993em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style=""><span class="mord mtight coloredeq eqe" style="">1</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span><span class="mopen" style="">(</span><span class="mord mathnormal" style="margin-right:0.04398em">z</span><span class="mclose" style="">)</span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">147</span> <span class="k">else</span><span class="p">:</span>
|
||||
<span class="lineno">148</span> <span class="k">return</span> <span class="o">-</span><span class="mi">1</span> <span class="o">*</span> <span class="bp">self</span><span class="o">.</span><span class="n">_terminal_utility_p1</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-29'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-29'>#</a>
|
||||
</div>
|
||||
<p>前两个事件是卡牌交易；即机会事件</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">150</span> <span class="k">def</span> <span class="nf">is_chance</span><span class="p">(</span><span class="bp">self</span><span class="p">)</span> <span class="o">-></span> <span class="nb">bool</span><span class="p">:</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-30'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-30'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">154</span> <span class="k">return</span> <span class="nb">len</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">history</span><span class="p">)</span> <span class="o"><</span> <span class="mi">2</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-31'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-31'>#</a>
|
||||
</div>
|
||||
<p>在历史记录中添加操作并返回新的历史记录</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">156</span> <span class="k">def</span> <span class="fm">__add__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">other</span><span class="p">:</span> <span class="n">Action</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-32'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-32'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">160</span> <span class="k">return</span> <span class="n">History</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">history</span> <span class="o">+</span> <span class="n">other</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-33'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-33'>#</a>
|
||||
</div>
|
||||
<p>当前玩家</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">162</span> <span class="k">def</span> <span class="nf">player</span><span class="p">(</span><span class="bp">self</span><span class="p">)</span> <span class="o">-></span> <span class="n">Player</span><span class="p">:</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-34'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-34'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">166</span> <span class="k">return</span> <span class="n">cast</span><span class="p">(</span><span class="n">Player</span><span class="p">,</span> <span class="nb">len</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">history</span><span class="p">)</span> <span class="o">%</span> <span class="mi">2</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-35'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-35'>#</a>
|
||||
</div>
|
||||
<p>采样一个机会动作</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">168</span> <span class="k">def</span> <span class="nf">sample_chance</span><span class="p">(</span><span class="bp">self</span><span class="p">)</span> <span class="o">-></span> <span class="n">Action</span><span class="p">:</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-36'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-36'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">172</span> <span class="k">while</span> <span class="kc">True</span><span class="p">:</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-37'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-37'>#</a>
|
||||
</div>
|
||||
<p>随机挑选一张牌</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">174</span> <span class="n">r</span> <span class="o">=</span> <span class="n">np</span><span class="o">.</span><span class="n">random</span><span class="o">.</span><span class="n">randint</span><span class="p">(</span><span class="nb">len</span><span class="p">(</span><span class="n">CHANCES</span><span class="p">))</span>
|
||||
<span class="lineno">175</span> <span class="n">chance</span> <span class="o">=</span> <span class="n">CHANCES</span><span class="p">[</span><span class="n">r</span><span class="p">]</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-38'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-38'>#</a>
|
||||
</div>
|
||||
<p>看看之前有没有发过牌</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">177</span> <span class="k">for</span> <span class="n">c</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">history</span><span class="p">:</span>
|
||||
<span class="lineno">178</span> <span class="k">if</span> <span class="n">c</span> <span class="o">==</span> <span class="n">chance</span><span class="p">:</span>
|
||||
<span class="lineno">179</span> <span class="n">chance</span> <span class="o">=</span> <span class="kc">None</span>
|
||||
<span class="lineno">180</span> <span class="k">break</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-39'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-39'>#</a>
|
||||
</div>
|
||||
<p>如果之前没有发过牌，则归还该牌</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">183</span> <span class="k">if</span> <span class="n">chance</span> <span class="ow">is</span> <span class="ow">not</span> <span class="kc">None</span><span class="p">:</span>
|
||||
<span class="lineno">184</span> <span class="k">return</span> <span class="n">cast</span><span class="p">(</span><span class="n">Action</span><span class="p">,</span> <span class="n">chance</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-40'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-40'>#</a>
|
||||
</div>
|
||||
<p>人类可读的表示</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">186</span> <span class="k">def</span> <span class="fm">__repr__</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-41'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-41'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">190</span> <span class="k">return</span> <span class="nb">repr</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">history</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-42'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-42'>#</a>
|
||||
</div>
|
||||
<p>当前历史记录的信息集关键字。这是一串仅对当前玩家可见的动作。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">192</span> <span class="k">def</span> <span class="nf">info_set_key</span><span class="p">(</span><span class="bp">self</span><span class="p">)</span> <span class="o">-></span> <span class="nb">str</span><span class="p">:</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-43'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-43'>#</a>
|
||||
</div>
|
||||
<p>获取当前玩家</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">198</span> <span class="n">i</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">player</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-44'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-44'>#</a>
|
||||
</div>
|
||||
<p>当前玩家看到她的牌和投注动作</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">200</span> <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">history</span><span class="p">[</span><span class="n">i</span><span class="p">]</span> <span class="o">+</span> <span class="bp">self</span><span class="o">.</span><span class="n">history</span><span class="p">[</span><span class="mi">2</span><span class="p">:]</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-45'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-45'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">202</span> <span class="k">def</span> <span class="nf">new_info_set</span><span class="p">(</span><span class="bp">self</span><span class="p">)</span> <span class="o">-></span> <span class="n">InfoSet</span><span class="p">:</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-46'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-46'>#</a>
|
||||
</div>
|
||||
<p>创建新的信息集对象</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">204</span> <span class="k">return</span> <span class="n">InfoSet</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">info_set_key</span><span class="p">())</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-47'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-47'>#</a>
|
||||
</div>
|
||||
<p>用于创建空历史对象的函数</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">207</span><span class="k">def</span> <span class="nf">create_new_history</span><span class="p">():</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-48'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-48'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">209</span> <span class="k">return</span> <span class="n">History</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-49'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-49'>#</a>
|
||||
</div>
|
||||
<p>配置扩展了 CFR 配置类</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">212</span><span class="k">class</span> <span class="nc">Configs</span><span class="p">(</span><span class="n">CFRConfigs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-50'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-50'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">216</span> <span class="k">pass</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-51'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-51'>#</a>
|
||||
</div>
|
||||
<p>设置库恩扑克<code class="highlight"><span></span><span class="n">create_new_history</span></code>
|
||||
的方法</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">219</span><span class="nd">@option</span><span class="p">(</span><span class="n">Configs</span><span class="o">.</span><span class="n">create_new_history</span><span class="p">)</span>
|
||||
<span class="lineno">220</span><span class="k">def</span> <span class="nf">_cnh</span><span class="p">():</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-52'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-52'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">224</span> <span class="k">return</span> <span class="n">create_new_history</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-53'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-53'>#</a>
|
||||
</div>
|
||||
<h3>运行实验</h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">227</span><span class="k">def</span> <span class="nf">main</span><span class="p">():</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-54'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-54'>#</a>
|
||||
</div>
|
||||
<p>创建一个实验，我们只写跟踪信息<code class="highlight"><span></span><span class="n">sqlite</span></code>
|
||||
来加快速度。由于算法的迭代速度很快，而且我们在每次迭代时都会跟踪数据，因此写入其他目标（如 Tensorboard）可能相对耗时。SQLite 对于我们的分析来说已经足够了。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">236</span> <span class="n">experiment</span><span class="o">.</span><span class="n">create</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">'kuhn_poker'</span><span class="p">,</span> <span class="n">writers</span><span class="o">=</span><span class="p">{</span><span class="s1">'sqlite'</span><span class="p">})</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-55'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-55'>#</a>
|
||||
</div>
|
||||
<p>初始化配置</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">238</span> <span class="n">conf</span> <span class="o">=</span> <span class="n">Configs</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-56'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-56'>#</a>
|
||||
</div>
|
||||
<p>装载配置</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">240</span> <span class="n">experiment</span><span class="o">.</span><span class="n">configs</span><span class="p">(</span><span class="n">conf</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-57'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-57'>#</a>
|
||||
</div>
|
||||
<p>开始实验</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">242</span> <span class="k">with</span> <span class="n">experiment</span><span class="o">.</span><span class="n">start</span><span class="p">():</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-58'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-58'>#</a>
|
||||
</div>
|
||||
<p>开始迭代</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">244</span> <span class="n">conf</span><span class="o">.</span><span class="n">cfr</span><span class="o">.</span><span class="n">iterate</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-59'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-59'>#</a>
|
||||
</div>
|
||||
<p></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">248</span><span class="k">if</span> <span class="vm">__name__</span> <span class="o">==</span> <span class="s1">'__main__'</span><span class="p">:</span>
|
||||
<span class="lineno">249</span> <span class="n">main</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
Binary file not shown.
|
After Width: | Height: | Size: 72 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 11 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 54 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 29 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 27 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 12 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 170 KiB |
@@ -0,0 +1,377 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content="在 CIFAR 10 上训练 ConvMixer"/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="在 CIFAR 10 上训练 ConvMixer"/>
|
||||
<meta name="twitter:description" content="在 CIFAR 10 上训练 ConvMixer"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/conv_mixer/experiment.html"/>
|
||||
<meta property="og:title" content="在 CIFAR 10 上训练 ConvMixer"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="在 CIFAR 10 上训练 ConvMixer"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="在 CIFAR 10 上训练 ConvMixer"/>
|
||||
<meta property="og:description" content="在 CIFAR 10 上训练 ConvMixer"/>
|
||||
|
||||
<title>在 CIFAR 10 上训练 ConvMixer</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/conv_mixer/experiment.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="index.html">conv_mixer</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/conv_mixer/experiment.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1>在 CIFA <a href="index.html">R 10 上训练 convMixer</a></h1>
|
||||
<p>此脚本在 CIFAR 10 数据集上训练 ConvMixer。</p>
|
||||
<p>这并不是试图重现论文的结果。本文使用 <a href="https://github.com/rwightman/pytorch-image-models">PyTorch 图像模型 (timm) 中存在的图像</a>增强进行训练。为了简单起见，我们没有这样做——这会导致我们的验证精度下降。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">18</span><span></span><span class="kn">from</span> <span class="nn">labml</span> <span class="kn">import</span> <span class="n">experiment</span>
|
||||
<span class="lineno">19</span><span class="kn">from</span> <span class="nn">labml.configs</span> <span class="kn">import</span> <span class="n">option</span>
|
||||
<span class="lineno">20</span><span class="kn">from</span> <span class="nn">labml_nn.experiments.cifar10</span> <span class="kn">import</span> <span class="n">CIFAR10Configs</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-1'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-1'>#</a>
|
||||
</div>
|
||||
<h2>配置</h2>
|
||||
<p>我们使用<a href="../experiments/cifar10.html"><code class="highlight"><span></span><span class="n">CIFAR10Configs</span></code>
|
||||
</a>它来定义所有与数据集相关的配置、优化器和训练循环。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">23</span><span class="k">class</span> <span class="nc">Configs</span><span class="p">(</span><span class="n">CIFAR10Configs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-2'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-2'>#</a>
|
||||
</div>
|
||||
<p>补丁的大小，<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.625em;vertical-align:-0.19444em;"></span><span class="mord mathnormal">p</span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">32</span> <span class="n">patch_size</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">2</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-3'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-3'>#</a>
|
||||
</div>
|
||||
<p>补丁嵌入中的通道数，<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.69444em;vertical-align:0em;"></span><span class="mord mathnormal">h</span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">34</span> <span class="n">d_model</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">256</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-4'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-4'>#</a>
|
||||
</div>
|
||||
<p><a href="#ConvMixerLayer">ConvMixer 层</a>数或深度，<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.69444em;vertical-align:0em;"></span><span class="mord mathnormal">d</span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">36</span> <span class="n">n_layers</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">8</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-5'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-5'>#</a>
|
||||
</div>
|
||||
<p>深度卷积的内核大小，<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.69444em;vertical-align:0em;"></span><span class="mord mathnormal" style="margin-right:0.03148em;">k</span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">38</span> <span class="n">kernel_size</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">7</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-6'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-6'>#</a>
|
||||
</div>
|
||||
<p>任务中的类数</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">40</span> <span class="n">n_classes</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">10</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-7'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-7'>#</a>
|
||||
</div>
|
||||
<h3>创建模型</h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">43</span><span class="nd">@option</span><span class="p">(</span><span class="n">Configs</span><span class="o">.</span><span class="n">model</span><span class="p">)</span>
|
||||
<span class="lineno">44</span><span class="k">def</span> <span class="nf">_conv_mixer</span><span class="p">(</span><span class="n">c</span><span class="p">:</span> <span class="n">Configs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-8'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-8'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">48</span> <span class="kn">from</span> <span class="nn">labml_nn.conv_mixer</span> <span class="kn">import</span> <span class="n">ConvMixerLayer</span><span class="p">,</span> <span class="n">ConvMixer</span><span class="p">,</span> <span class="n">ClassificationHead</span><span class="p">,</span> <span class="n">PatchEmbeddings</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-9'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-9'>#</a>
|
||||
</div>
|
||||
<p>创建混音器</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">51</span> <span class="k">return</span> <span class="n">ConvMixer</span><span class="p">(</span><span class="n">ConvMixerLayer</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">d_model</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">kernel_size</span><span class="p">),</span> <span class="n">c</span><span class="o">.</span><span class="n">n_layers</span><span class="p">,</span>
|
||||
<span class="lineno">52</span> <span class="n">PatchEmbeddings</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">d_model</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">patch_size</span><span class="p">,</span> <span class="mi">3</span><span class="p">),</span>
|
||||
<span class="lineno">53</span> <span class="n">ClassificationHead</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">d_model</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">n_classes</span><span class="p">))</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">device</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-10'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-10'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">56</span><span class="k">def</span> <span class="nf">main</span><span class="p">():</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-11'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-11'>#</a>
|
||||
</div>
|
||||
<p>创建实验</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">58</span> <span class="n">experiment</span><span class="o">.</span><span class="n">create</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">'ConvMixer'</span><span class="p">,</span> <span class="n">comment</span><span class="o">=</span><span class="s1">'cifar10'</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-12'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-12'>#</a>
|
||||
</div>
|
||||
<p>创建配置</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">60</span> <span class="n">conf</span> <span class="o">=</span> <span class="n">Configs</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-13'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-13'>#</a>
|
||||
</div>
|
||||
<p>装载配置</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">62</span> <span class="n">experiment</span><span class="o">.</span><span class="n">configs</span><span class="p">(</span><span class="n">conf</span><span class="p">,</span> <span class="p">{</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-14'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-14'>#</a>
|
||||
</div>
|
||||
<p>优化器</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">64</span> <span class="s1">'optimizer.optimizer'</span><span class="p">:</span> <span class="s1">'Adam'</span><span class="p">,</span>
|
||||
<span class="lineno">65</span> <span class="s1">'optimizer.learning_rate'</span><span class="p">:</span> <span class="mf">2.5e-4</span><span class="p">,</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-15'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-15'>#</a>
|
||||
</div>
|
||||
<p>训练周期和批次大小</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">68</span> <span class="s1">'epochs'</span><span class="p">:</span> <span class="mi">150</span><span class="p">,</span>
|
||||
<span class="lineno">69</span> <span class="s1">'train_batch_size'</span><span class="p">:</span> <span class="mi">64</span><span class="p">,</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-16'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-16'>#</a>
|
||||
</div>
|
||||
<p>简单的图像增强</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">72</span> <span class="s1">'train_dataset'</span><span class="p">:</span> <span class="s1">'cifar10_train_augmented'</span><span class="p">,</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-17'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-17'>#</a>
|
||||
</div>
|
||||
<p>不要扩充图像以进行验证</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">74</span> <span class="s1">'valid_dataset'</span><span class="p">:</span> <span class="s1">'cifar10_valid_no_augment'</span><span class="p">,</span>
|
||||
<span class="lineno">75</span> <span class="p">})</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-18'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-18'>#</a>
|
||||
</div>
|
||||
<p>设置保存/加载的模型</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">77</span> <span class="n">experiment</span><span class="o">.</span><span class="n">add_pytorch_models</span><span class="p">({</span><span class="s1">'model'</span><span class="p">:</span> <span class="n">conf</span><span class="o">.</span><span class="n">model</span><span class="p">})</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-19'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-19'>#</a>
|
||||
</div>
|
||||
<p>开始实验并运行训练循环</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">79</span> <span class="k">with</span> <span class="n">experiment</span><span class="o">.</span><span class="n">start</span><span class="p">():</span>
|
||||
<span class="lineno">80</span> <span class="n">conf</span><span class="o">.</span><span class="n">run</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-20'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-20'>#</a>
|
||||
</div>
|
||||
<p></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">84</span><span class="k">if</span> <span class="vm">__name__</span> <span class="o">==</span> <span class="s1">'__main__'</span><span class="p">:</span>
|
||||
<span class="lineno">85</span> <span class="n">main</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
@@ -0,0 +1,725 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content="论文 “补丁就是你所需要的吗？” 的 PyTorch 实现/教程"/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="补丁是你所需要的吗？（convMixer）"/>
|
||||
<meta name="twitter:description" content="论文 “补丁就是你所需要的吗？” 的 PyTorch 实现/教程"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/conv_mixer/index.html"/>
|
||||
<meta property="og:title" content="补丁是你所需要的吗？（convMixer）"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="补丁是你所需要的吗？（convMixer）"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="补丁是你所需要的吗？（convMixer）"/>
|
||||
<meta property="og:description" content="论文 “补丁就是你所需要的吗？” 的 PyTorch 实现/教程"/>
|
||||
|
||||
<title>补丁是你所需要的吗？（convMixer）</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/conv_mixer/index.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="index.html">conv_mixer</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/conv_mixer/__init__.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1>你只需要补丁吗？（convMixer）</h1>
|
||||
<p>这是 <a href="https://pytorch.org">PyTorch</a> 对论文《<a href="https://arxiv.org/abs/2201.09792">补丁就是你所需要的？</a>》的实现</p>。
|
||||
<p><img alt="ConvMixer diagram from the paper" src="conv_mixer.png"></p>
|
||||
<p>convMixer 类似于 <a href="../transformers/mlp_mixer/index.html">MLP 混音器</a>。MLP-Mixer 通过在空间维度上应用 MLP，然后在信道维度上应用 MLP 来分离空间维度和信道维度的混音（空间 MLP 取代 <a href="../transformers/vit/index.html">vIT</a> 注意力，信道 MLP 是 ViT 的 <a href="../transformers/feed_forward.html">FFN</a>）。</p>
|
||||
<p>ConvMixer 使用<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.72777em;vertical-align:-0.08333em;"></span><span class="mord coloredeq eqa" style=""><span class="mord" style="">1</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">×</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord" style="">1</span></span></span></span></span></span>卷积进行通道混合，使用深度卷积进行空间混合。由于它是卷积而不是整个空间的完整的 MLP，因此与 vIT 或 MLP-Mixer 相比，它只混合附近的批次。此外，MLP-Mixer 每次混合使用两层 MLP，ConvMixer 每次混合使用单层。</p>
|
||||
<p>该论文建议删除信道混合（逐点卷积）上的剩余连接，在空间混合（深度卷积）上仅使用残差连接。他们还使用<a href="../normalization/batch_norm/index.html">批量标准化</a>而不是<a href="../normalization/layer_norm/index.html">图层标准化</a>。</p>
|
||||
<p>这是<a href="experiment.html">一项在 CIFAR-10 上训练 ConvMixer 的实验</a>。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">36</span><span></span><span class="kn">import</span> <span class="nn">torch</span>
|
||||
<span class="lineno">37</span><span class="kn">from</span> <span class="nn">torch</span> <span class="kn">import</span> <span class="n">nn</span>
|
||||
<span class="lineno">38</span>
|
||||
<span class="lineno">39</span><span class="kn">from</span> <span class="nn">labml_helpers.module</span> <span class="kn">import</span> <span class="n">Module</span>
|
||||
<span class="lineno">40</span><span class="kn">from</span> <span class="nn">labml_nn.utils</span> <span class="kn">import</span> <span class="n">clone_module_list</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-1'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-1'>#</a>
|
||||
</div>
|
||||
<p><a id="ConvMixerLayer"></a></p>
|
||||
<h2>混音器层</h2>
|
||||
<p>这是单个 ConvMixer 层。该模型将有一系列这样的。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">43</span><span class="k">class</span> <span class="nc">ConvMixerLayer</span><span class="p">(</span><span class="n">Module</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-2'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-2'>#</a>
|
||||
</div>
|
||||
<ul><li><code class="highlight"><span></span><span class="n">d_model</span></code>
|
||||
是补丁嵌入中的通道数，<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.69444em;vertical-align:0em;"></span><span class="mord coloredeq eqd" style=""><span class="mord mathnormal" style="">h</span></span></span></span></span></span></li>
|
||||
<li><code class="highlight"><span></span><span class="n">kernel_size</span></code>
|
||||
是空间卷积内核的大小，<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.69444em;vertical-align:0em;"></span><span class="mord mathnormal" style="margin-right:0.03148em;">k</span></span></span></span></span></li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">52</span> <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">d_model</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">kernel_size</span><span class="p">:</span> <span class="nb">int</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-3'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-3'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">57</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-4'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-4'>#</a>
|
||||
</div>
|
||||
<p>深度卷积是每个通道的单独卷积。我们使用卷积层来完成此操作，该卷积层的组数等于通道数。因此，每个频道都是它自己的组。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">61</span> <span class="bp">self</span><span class="o">.</span><span class="n">depth_wise_conv</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Conv2d</span><span class="p">(</span><span class="n">d_model</span><span class="p">,</span> <span class="n">d_model</span><span class="p">,</span>
|
||||
<span class="lineno">62</span> <span class="n">kernel_size</span><span class="o">=</span><span class="n">kernel_size</span><span class="p">,</span>
|
||||
<span class="lineno">63</span> <span class="n">groups</span><span class="o">=</span><span class="n">d_model</span><span class="p">,</span>
|
||||
<span class="lineno">64</span> <span class="n">padding</span><span class="o">=</span><span class="p">(</span><span class="n">kernel_size</span> <span class="o">-</span> <span class="mi">1</span><span class="p">)</span> <span class="o">//</span> <span class="mi">2</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-5'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-5'>#</a>
|
||||
</div>
|
||||
<p>深度卷积后激活</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">66</span> <span class="bp">self</span><span class="o">.</span><span class="n">act1</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">GELU</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-6'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-6'>#</a>
|
||||
</div>
|
||||
<p>深度卷积后的归一化</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">68</span> <span class="bp">self</span><span class="o">.</span><span class="n">norm1</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">BatchNorm2d</span><span class="p">(</span><span class="n">d_model</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-7'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-7'>#</a>
|
||||
</div>
|
||||
<p>逐点卷积是一种<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.72777em;vertical-align:-0.08333em;"></span><span class="mord coloredeq eqa" style=""><span class="mord" style="">1</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">×</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord" style="">1</span></span></span></span></span></span>卷积。即补丁嵌入的线性变换</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">72</span> <span class="bp">self</span><span class="o">.</span><span class="n">point_wise_conv</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Conv2d</span><span class="p">(</span><span class="n">d_model</span><span class="p">,</span> <span class="n">d_model</span><span class="p">,</span> <span class="n">kernel_size</span><span class="o">=</span><span class="mi">1</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-8'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-8'>#</a>
|
||||
</div>
|
||||
<p>逐点卷积后激活</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">74</span> <span class="bp">self</span><span class="o">.</span><span class="n">act2</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">GELU</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-9'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-9'>#</a>
|
||||
</div>
|
||||
<p>逐点卷积后的归一化</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">76</span> <span class="bp">self</span><span class="o">.</span><span class="n">norm2</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">BatchNorm2d</span><span class="p">(</span><span class="n">d_model</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-10'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-10'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">78</span> <span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-11'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-11'>#</a>
|
||||
</div>
|
||||
<p>对于围绕深度卷积的剩余连接</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">80</span> <span class="n">residual</span> <span class="o">=</span> <span class="n">x</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-12'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-12'>#</a>
|
||||
</div>
|
||||
<p>深度卷积、激活和归一化</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">83</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">depth_wise_conv</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
|
||||
<span class="lineno">84</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">act1</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
|
||||
<span class="lineno">85</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">norm1</span><span class="p">(</span><span class="n">x</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-13'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-13'>#</a>
|
||||
</div>
|
||||
<p>添加剩余连接</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">88</span> <span class="n">x</span> <span class="o">+=</span> <span class="n">residual</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-14'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-14'>#</a>
|
||||
</div>
|
||||
<p>逐点卷积、激活和归一化</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">91</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">point_wise_conv</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
|
||||
<span class="lineno">92</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">act2</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
|
||||
<span class="lineno">93</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">norm2</span><span class="p">(</span><span class="n">x</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-15'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-15'>#</a>
|
||||
</div>
|
||||
<p></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">96</span> <span class="k">return</span> <span class="n">x</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-16'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-16'>#</a>
|
||||
</div>
|
||||
<p><a id="PatchEmbeddings"></a></p>
|
||||
<h2>获取补丁嵌入</h2>
|
||||
<p>这会将图像拆分为大小的补丁，<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.7777700000000001em;vertical-align:-0.19444em;"></span><span class="mord coloredeq eqf" style=""><span class="mord mathnormal" style="">p</span></span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin">×</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span></span><span class="base"><span class="strut" style="height:0.625em;vertical-align:-0.19444em;"></span><span class="mord coloredeq eqf" style=""><span class="mord mathnormal" style="">p</span></span></span></span></span></span>并为每个补丁提供嵌入。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">99</span><span class="k">class</span> <span class="nc">PatchEmbeddings</span><span class="p">(</span><span class="n">Module</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-17'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-17'>#</a>
|
||||
</div>
|
||||
<ul><li><code class="highlight"><span></span><span class="n">d_model</span></code>
|
||||
是补丁嵌入中的通道数<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.69444em;vertical-align:0em;"></span><span class="mord coloredeq eqd" style=""><span class="mord mathnormal" style="">h</span></span></span></span></span></span></li>
|
||||
<li><code class="highlight"><span></span><span class="n">patch_size</span></code>
|
||||
是补丁的大小，<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.625em;vertical-align:-0.19444em;"></span><span class="mord coloredeq eqf" style=""><span class="mord mathnormal" style="">p</span></span></span></span></span></span></li>
|
||||
<li><code class="highlight"><span></span><span class="n">in_channels</span></code>
|
||||
是输入图像中的通道数（rgb 为 3）</li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">108</span> <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">d_model</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">patch_size</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">in_channels</span><span class="p">:</span> <span class="nb">int</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-18'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-18'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">114</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-19'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-19'>#</a>
|
||||
</div>
|
||||
<p>我们创建一个卷积层，其内核大小和步长等于补丁大小。这相当于将图像分割成色块并在每个面片上进行线性变换。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">119</span> <span class="bp">self</span><span class="o">.</span><span class="n">conv</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Conv2d</span><span class="p">(</span><span class="n">in_channels</span><span class="p">,</span> <span class="n">d_model</span><span class="p">,</span> <span class="n">kernel_size</span><span class="o">=</span><span class="n">patch_size</span><span class="p">,</span> <span class="n">stride</span><span class="o">=</span><span class="n">patch_size</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-20'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-20'>#</a>
|
||||
</div>
|
||||
<p>激活功能</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">121</span> <span class="bp">self</span><span class="o">.</span><span class="n">act</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">GELU</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-21'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-21'>#</a>
|
||||
</div>
|
||||
<p>批量标准化</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">123</span> <span class="bp">self</span><span class="o">.</span><span class="n">norm</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">BatchNorm2d</span><span class="p">(</span><span class="n">d_model</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-22'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-22'>#</a>
|
||||
</div>
|
||||
<ul><li><code class="highlight"><span></span><span class="n">x</span></code>
|
||||
是形状的输入图像<code class="highlight"><span></span><span class="p">[</span><span class="n">batch_size</span><span class="p">,</span> <span class="n">channels</span><span class="p">,</span> <span class="n">height</span><span class="p">,</span> <span class="n">width</span><span class="p">]</span></code>
|
||||
</li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">125</span> <span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-23'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-23'>#</a>
|
||||
</div>
|
||||
<p>应用卷积层</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">130</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">conv</span><span class="p">(</span><span class="n">x</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-24'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-24'>#</a>
|
||||
</div>
|
||||
<p>激活和规范化</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">132</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">act</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
|
||||
<span class="lineno">133</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">norm</span><span class="p">(</span><span class="n">x</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-25'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-25'>#</a>
|
||||
</div>
|
||||
<p></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">136</span> <span class="k">return</span> <span class="n">x</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-26'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-26'>#</a>
|
||||
</div>
|
||||
<p><a id="ClassificationHead"></a></p>
|
||||
<h2>分类主管</h2>
|
||||
<p>它们进行平均池（取所有补丁嵌入的均值）和最终的线性变换来预测影像类的对数概率。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">139</span><span class="k">class</span> <span class="nc">ClassificationHead</span><span class="p">(</span><span class="n">Module</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-27'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-27'>#</a>
|
||||
</div>
|
||||
<ul><li><code class="highlight"><span></span><span class="n">d_model</span></code>
|
||||
是补丁嵌入中的通道数，<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.69444em;vertical-align:0em;"></span><span class="mord coloredeq eqd" style=""><span class="mord mathnormal" style="">h</span></span></span></span></span></span></li>
|
||||
<li><code class="highlight"><span></span><span class="n">n_classes</span></code>
|
||||
是分类任务中的类数</li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">149</span> <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">d_model</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">n_classes</span><span class="p">:</span> <span class="nb">int</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-28'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-28'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">154</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-29'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-29'>#</a>
|
||||
</div>
|
||||
<p>平均池</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">156</span> <span class="bp">self</span><span class="o">.</span><span class="n">pool</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">AdaptiveAvgPool2d</span><span class="p">((</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">))</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-30'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-30'>#</a>
|
||||
</div>
|
||||
<p>线性层</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">158</span> <span class="bp">self</span><span class="o">.</span><span class="n">linear</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Linear</span><span class="p">(</span><span class="n">d_model</span><span class="p">,</span> <span class="n">n_classes</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-31'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-31'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">160</span> <span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-32'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-32'>#</a>
|
||||
</div>
|
||||
<p>平均汇集</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">162</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">pool</span><span class="p">(</span><span class="n">x</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-33'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-33'>#</a>
|
||||
</div>
|
||||
<p>得到嵌入，<code class="highlight"><span></span><span class="n">x</span></code>
|
||||
会有形状<code class="highlight"><span></span><span class="p">[</span><span class="n">batch_size</span><span class="p">,</span> <span class="n">d_model</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">]</span></code>
|
||||
</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">164</span> <span class="n">x</span> <span class="o">=</span> <span class="n">x</span><span class="p">[:,</span> <span class="p">:,</span> <span class="mi">0</span><span class="p">,</span> <span class="mi">0</span><span class="p">]</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-34'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-34'>#</a>
|
||||
</div>
|
||||
<p>线性层</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">166</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">linear</span><span class="p">(</span><span class="n">x</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-35'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-35'>#</a>
|
||||
</div>
|
||||
<p></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">169</span> <span class="k">return</span> <span class="n">x</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-36'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-36'>#</a>
|
||||
</div>
|
||||
<h2>混音器</h2>
|
||||
<p>它结合了补丁嵌入块、许多 ConvMixer 层和一个分类头。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">172</span><span class="k">class</span> <span class="nc">ConvMixer</span><span class="p">(</span><span class="n">Module</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-37'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-37'>#</a>
|
||||
</div>
|
||||
<ul><li><code class="highlight"><span></span><span class="n">conv_mixer_layer</span></code>
|
||||
是单个 C <a href="#ConvMixerLayer">onvMixer 层</a>的副本。我们制作它的副本来制作 ConvMixer<code class="highlight"><span></span><span class="n">n_layers</span></code>
|
||||
。</li>
|
||||
<li><code class="highlight"><span></span><span class="n">n_layers</span></code>
|
||||
是 ConvMixer 层（或深度）的数量<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.69444em;vertical-align:0em;"></span><span class="mord mathnormal">d</span></span></span></span></span>。</li>
|
||||
<li><code class="highlight"><span></span><span class="n">patch_emb</span></code>
|
||||
是<a href="#PatchEmbeddings">补丁嵌入层</a>。</li>
|
||||
<li><code class="highlight"><span></span><span class="n">classification</span></code>
|
||||
是<a href="#ClassificationHead">分类头</a>。</li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">179</span> <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">conv_mixer_layer</span><span class="p">:</span> <span class="n">ConvMixerLayer</span><span class="p">,</span> <span class="n">n_layers</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span>
|
||||
<span class="lineno">180</span> <span class="n">patch_emb</span><span class="p">:</span> <span class="n">PatchEmbeddings</span><span class="p">,</span>
|
||||
<span class="lineno">181</span> <span class="n">classification</span><span class="p">:</span> <span class="n">ClassificationHead</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-38'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-38'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">189</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-39'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-39'>#</a>
|
||||
</div>
|
||||
<p>补丁嵌入</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">191</span> <span class="bp">self</span><span class="o">.</span><span class="n">patch_emb</span> <span class="o">=</span> <span class="n">patch_emb</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-40'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-40'>#</a>
|
||||
</div>
|
||||
<p>分类主管</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">193</span> <span class="bp">self</span><span class="o">.</span><span class="n">classification</span> <span class="o">=</span> <span class="n">classification</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-41'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-41'>#</a>
|
||||
</div>
|
||||
<p>制作 C <a href="#ConvMixerLayer">onvMixer 图层</a>的副本</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">195</span> <span class="bp">self</span><span class="o">.</span><span class="n">conv_mixer_layers</span> <span class="o">=</span> <span class="n">clone_module_list</span><span class="p">(</span><span class="n">conv_mixer_layer</span><span class="p">,</span> <span class="n">n_layers</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-42'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-42'>#</a>
|
||||
</div>
|
||||
<ul><li><code class="highlight"><span></span><span class="n">x</span></code>
|
||||
是形状的输入图像<code class="highlight"><span></span><span class="p">[</span><span class="n">batch_size</span><span class="p">,</span> <span class="n">channels</span><span class="p">,</span> <span class="n">height</span><span class="p">,</span> <span class="n">width</span><span class="p">]</span></code>
|
||||
</li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">197</span> <span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-43'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-43'>#</a>
|
||||
</div>
|
||||
<p>获取补丁嵌入。这给出了形状的张量<code class="highlight"><span></span><span class="p">[</span><span class="n">batch_size</span><span class="p">,</span> <span class="n">d_model</span><span class="p">,</span> <span class="n">height</span> <span class="o">/</span> <span class="n">patch_size</span><span class="p">,</span> <span class="n">width</span> <span class="o">/</span> <span class="n">patch_size</span><span class="p">]</span></code>
|
||||
。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">202</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">patch_emb</span><span class="p">(</span><span class="n">x</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-44'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-44'>#</a>
|
||||
</div>
|
||||
<p>穿过 <a href="#ConvMixerLayer">ConvMixer 图层</a></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">205</span> <span class="k">for</span> <span class="n">layer</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">conv_mixer_layers</span><span class="p">:</span>
|
||||
<span class="lineno">206</span> <span class="n">x</span> <span class="o">=</span> <span class="n">layer</span><span class="p">(</span><span class="n">x</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-45'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-45'>#</a>
|
||||
</div>
|
||||
<p>分类头，获取日志</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">209</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">classification</span><span class="p">(</span><span class="n">x</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-46'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-46'>#</a>
|
||||
</div>
|
||||
<p></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">212</span> <span class="k">return</span> <span class="n">x</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
@@ -0,0 +1,130 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content=""/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content=" 补丁是你所需要的吗？"/>
|
||||
<meta name="twitter:description" content=""/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/conv_mixer/readme.html"/>
|
||||
<meta property="og:title" content=" 补丁是你所需要的吗？"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content=" 补丁是你所需要的吗？"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content=" 补丁是你所需要的吗？"/>
|
||||
<meta property="og:description" content=""/>
|
||||
|
||||
<title> 补丁是你所需要的吗？</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/conv_mixer/readme.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="index.html">conv_mixer</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/conv_mixer/readme.md" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1><a href="https://nn.labml.ai/conv_mixer/index.html">你只需要补丁吗？</a></h1>
|
||||
<p>这是 <a href="https://pytorch.org">PyTorch</a> 对论文《<a href="https://arxiv.org/abs/2201.09792">补丁就是你所需要的？</a>》的实现</p>。
|
||||
<p>convMixer 类似于 <a href="https://nn.labml.ai/transformers/mlp_mixer/index.html">MLP 混音器</a>。MLP-Mixer 通过在空间维度上应用 MLP，然后在信道维度上应用 MLP 来分离空间维度和信道维度的混音（空间 MLP 取代 <a href="https://nn.labml.ai/transformers/vit/index.html">vIT</a> 注意力，信道 MLP 是 ViT 的 <a href="https://nn.labml.ai/transformers/feed_forward.html">FFN</a>）。</p>
|
||||
<p>ConvMixer 使用 1x1 卷积进行通道混合，使用深度卷积进行空间混合。由于它是卷积而不是整个空间的完整的 MLP，因此与 vIT 或 MLP-Mixer 相比，它只混合附近的批次。此外，MLP-Mixer 每次混合使用两层 MLP，ConvMixer 每次混合使用单层。</p>
|
||||
<p>该论文建议删除信道混合（逐点卷积）上的剩余连接，在空间混合（深度卷积）上仅使用残差连接。他们还使用<a href="https://nn.labml.ai/normalization/batch_norm/index.html">批量标准化</a>而不是<a href="../normalization/layer_norm/index.html">图层标准化</a>。</p>
|
||||
<p>这是<a href="https://nn.labml.ai/conv_mixer/experiment.html">一项在 CIFAR-10 上训练 ConvMixer 的实验</a>。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
File diff suppressed because one or more lines are too long
@@ -0,0 +1,990 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content="降噪扩散概率模型的训练代码。"/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="去噪扩散概率模型 (DDPM) 训练"/>
|
||||
<meta name="twitter:description" content="降噪扩散概率模型的训练代码。"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/diffusion/ddpm/experiment.html"/>
|
||||
<meta property="og:title" content="去噪扩散概率模型 (DDPM) 训练"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="去噪扩散概率模型 (DDPM) 训练"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="去噪扩散概率模型 (DDPM) 训练"/>
|
||||
<meta property="og:description" content="降噪扩散概率模型的训练代码。"/>
|
||||
|
||||
<title>去噪扩散概率模型 (DDPM) 训练</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/diffusion/ddpm/experiment.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="../index.html">diffusion</a>
|
||||
<a class="parent" href="index.html">ddpm</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/diffusion/ddpm/experiment.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1><a href="index.html">降噪扩散概率模型 (DDPM)</a> 训练</h1>
|
||||
<p><a href="https://colab.research.google.com/github/labmlai/annotated_deep_learning_paper_implementations/blob/master/labml_nn/diffusion/ddpm/experiment.ipynb"><img alt="Open In Colab" src="https://colab.research.google.com/assets/colab-badge.svg"></a></p>
|
||||
<p>这将基于 CeleBA HQ 数据集训练基于 DDPM 的模型。你可以在 <a href="https://forums.fast.ai/t/download-celeba-hq-dataset/45873/3">fast.ai 的讨论</a>中找到下载说明。将图像保存在<a href="#dataset_path"><code class="highlight"><span></span><span class="n">data</span><span class="o">/</span><span class="n">celebA</span></code>
|
||||
文件夹中</a>。</p>
|
||||
<p>该论文使用了该模型的指数移动平均线，其衰减量为<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord">0.9999</span></span></span></span></span>。为简单起见，我们跳过了这个。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">20</span><span></span><span class="kn">from</span> <span class="nn">typing</span> <span class="kn">import</span> <span class="n">List</span>
|
||||
<span class="lineno">21</span>
|
||||
<span class="lineno">22</span><span class="kn">import</span> <span class="nn">torch</span>
|
||||
<span class="lineno">23</span><span class="kn">import</span> <span class="nn">torch.utils.data</span>
|
||||
<span class="lineno">24</span><span class="kn">import</span> <span class="nn">torchvision</span>
|
||||
<span class="lineno">25</span><span class="kn">from</span> <span class="nn">PIL</span> <span class="kn">import</span> <span class="n">Image</span>
|
||||
<span class="lineno">26</span>
|
||||
<span class="lineno">27</span><span class="kn">from</span> <span class="nn">labml</span> <span class="kn">import</span> <span class="n">lab</span><span class="p">,</span> <span class="n">tracker</span><span class="p">,</span> <span class="n">experiment</span><span class="p">,</span> <span class="n">monit</span>
|
||||
<span class="lineno">28</span><span class="kn">from</span> <span class="nn">labml.configs</span> <span class="kn">import</span> <span class="n">BaseConfigs</span><span class="p">,</span> <span class="n">option</span>
|
||||
<span class="lineno">29</span><span class="kn">from</span> <span class="nn">labml_helpers.device</span> <span class="kn">import</span> <span class="n">DeviceConfigs</span>
|
||||
<span class="lineno">30</span><span class="kn">from</span> <span class="nn">labml_nn.diffusion.ddpm</span> <span class="kn">import</span> <span class="n">DenoiseDiffusion</span>
|
||||
<span class="lineno">31</span><span class="kn">from</span> <span class="nn">labml_nn.diffusion.ddpm.unet</span> <span class="kn">import</span> <span class="n">UNet</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-1'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-1'>#</a>
|
||||
</div>
|
||||
<h2>配置</h2>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">34</span><span class="k">class</span> <span class="nc">Configs</span><span class="p">(</span><span class="n">BaseConfigs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-2'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-2'>#</a>
|
||||
</div>
|
||||
<p>用于训练模型的设备。<a href="https://docs.labml.ai/api/helpers.html#labml_helpers.device.DeviceConfigs"><code class="highlight"><span></span><span class="n">DeviceConfigs</span></code>
|
||||
</a>选择可用的 CUDA 设备或默认为 CPU。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">41</span> <span class="n">device</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">device</span> <span class="o">=</span> <span class="n">DeviceConfigs</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-3'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-3'>#</a>
|
||||
</div>
|
||||
<p>U-Net 模型用于<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:1em;vertical-align:-0.25em;"></span><span class="mord"><span class="mord" style="color:lightgreen"><span class="mord mathnormal" style="">ϵ</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.33610799999999996em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mathnormal mtight" style="margin-right:0.02778em">θ</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span><span class="mopen coloredeq eqb" style="">(</span><span class="mord coloredeq eqb" style=""><span class="mord mathnormal" style="">x</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.2805559999999999em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style=""><span class="mord mathnormal mtight coloredeq eqg" style="">t</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span><span class="mpunct coloredeq eqb" style="">,</span><span class="mspace" style="margin-right:0.16666666666666666em;"></span><span class="mord coloredeq eqb" style=""><span class="mord mathnormal coloredeq eqg" style="">t</span></span><span class="mclose coloredeq eqb" style="">)</span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">44</span> <span class="n">eps_model</span><span class="p">:</span> <span class="n">UNet</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-4'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-4'>#</a>
|
||||
</div>
|
||||
<p><a href="index.html">DDPM 算法</a></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">46</span> <span class="n">diffusion</span><span class="p">:</span> <span class="n">DenoiseDiffusion</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-5'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-5'>#</a>
|
||||
</div>
|
||||
<p>图像中的通道数。<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord">3</span></span></span></span></span>对于 RGB。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">49</span> <span class="n">image_channels</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">3</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-6'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-6'>#</a>
|
||||
</div>
|
||||
<p>图像大小</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">51</span> <span class="n">image_size</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">32</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-7'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-7'>#</a>
|
||||
</div>
|
||||
<p>初始特征图中的频道数量</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">53</span> <span class="n">n_channels</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">64</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-8'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-8'>#</a>
|
||||
</div>
|
||||
<p>每种分辨率下的通道编号列表。频道的数量是<code class="highlight"><span></span><span class="n">channel_multipliers</span><span class="p">[</span><span class="n">i</span><span class="p">]</span> <span class="o">*</span> <span class="n">n_channels</span></code>
|
||||
</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">56</span> <span class="n">channel_multipliers</span><span class="p">:</span> <span class="n">List</span><span class="p">[</span><span class="nb">int</span><span class="p">]</span> <span class="o">=</span> <span class="p">[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="mi">4</span><span class="p">]</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-9'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-9'>#</a>
|
||||
</div>
|
||||
<p>指示是否在每个分辨率下使用注意力的布尔值列表</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">58</span> <span class="n">is_attention</span><span class="p">:</span> <span class="n">List</span><span class="p">[</span><span class="nb">int</span><span class="p">]</span> <span class="o">=</span> <span class="p">[</span><span class="kc">False</span><span class="p">,</span> <span class="kc">False</span><span class="p">,</span> <span class="kc">False</span><span class="p">,</span> <span class="kc">True</span><span class="p">]</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-10'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-10'>#</a>
|
||||
</div>
|
||||
<p>时间步数<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.68333em;vertical-align:0em;"></span><span class="mord coloredeq eqf" style=""><span class="mord mathnormal" style="margin-right:0.13889em">T</span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">61</span> <span class="n">n_steps</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">1_000</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-11'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-11'>#</a>
|
||||
</div>
|
||||
<p>批量大小</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">63</span> <span class="n">batch_size</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">64</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-12'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-12'>#</a>
|
||||
</div>
|
||||
<p>要生成的样本数</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">65</span> <span class="n">n_samples</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">16</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-13'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-13'>#</a>
|
||||
</div>
|
||||
<p>学习率</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">67</span> <span class="n">learning_rate</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">2e-5</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-14'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-14'>#</a>
|
||||
</div>
|
||||
<p>训练周期的数量</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">70</span> <span class="n">epochs</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">1_000</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-15'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-15'>#</a>
|
||||
</div>
|
||||
<p>数据集</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">73</span> <span class="n">dataset</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">utils</span><span class="o">.</span><span class="n">data</span><span class="o">.</span><span class="n">Dataset</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-16'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-16'>#</a>
|
||||
</div>
|
||||
<p>数据加载器</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">75</span> <span class="n">data_loader</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">utils</span><span class="o">.</span><span class="n">data</span><span class="o">.</span><span class="n">DataLoader</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-17'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-17'>#</a>
|
||||
</div>
|
||||
<p>Adam 优化器</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">78</span> <span class="n">optimizer</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">optim</span><span class="o">.</span><span class="n">Adam</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-18'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-18'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">80</span> <span class="k">def</span> <span class="nf">init</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-19'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-19'>#</a>
|
||||
</div>
|
||||
<p>创建<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:1em;vertical-align:-0.25em;"></span><span class="mord"><span class="mord" style="color:lightgreen"><span class="mord mathnormal" style="">ϵ</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.33610799999999996em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mathnormal mtight" style="margin-right:0.02778em">θ</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span><span class="mopen coloredeq eqb" style="">(</span><span class="mord coloredeq eqb" style=""><span class="mord mathnormal" style="">x</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.2805559999999999em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style=""><span class="mord mathnormal mtight coloredeq eqg" style="">t</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span><span class="mpunct coloredeq eqb" style="">,</span><span class="mspace" style="margin-right:0.16666666666666666em;"></span><span class="mord coloredeq eqb" style=""><span class="mord mathnormal coloredeq eqg" style="">t</span></span><span class="mclose coloredeq eqb" style="">)</span></span></span></span></span></span>模型</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">82</span> <span class="bp">self</span><span class="o">.</span><span class="n">eps_model</span> <span class="o">=</span> <span class="n">UNet</span><span class="p">(</span>
|
||||
<span class="lineno">83</span> <span class="n">image_channels</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">image_channels</span><span class="p">,</span>
|
||||
<span class="lineno">84</span> <span class="n">n_channels</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">n_channels</span><span class="p">,</span>
|
||||
<span class="lineno">85</span> <span class="n">ch_mults</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">channel_multipliers</span><span class="p">,</span>
|
||||
<span class="lineno">86</span> <span class="n">is_attn</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">is_attention</span><span class="p">,</span>
|
||||
<span class="lineno">87</span> <span class="p">)</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">device</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-20'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-20'>#</a>
|
||||
</div>
|
||||
<p>创建 <a href="index.html">DDPM 类</a></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">90</span> <span class="bp">self</span><span class="o">.</span><span class="n">diffusion</span> <span class="o">=</span> <span class="n">DenoiseDiffusion</span><span class="p">(</span>
|
||||
<span class="lineno">91</span> <span class="n">eps_model</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">eps_model</span><span class="p">,</span>
|
||||
<span class="lineno">92</span> <span class="n">n_steps</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">n_steps</span><span class="p">,</span>
|
||||
<span class="lineno">93</span> <span class="n">device</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">device</span><span class="p">,</span>
|
||||
<span class="lineno">94</span> <span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-21'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-21'>#</a>
|
||||
</div>
|
||||
<p>创建数据加载器</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">97</span> <span class="bp">self</span><span class="o">.</span><span class="n">data_loader</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">utils</span><span class="o">.</span><span class="n">data</span><span class="o">.</span><span class="n">DataLoader</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">dataset</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">batch_size</span><span class="p">,</span> <span class="n">shuffle</span><span class="o">=</span><span class="kc">True</span><span class="p">,</span> <span class="n">pin_memory</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-22'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-22'>#</a>
|
||||
</div>
|
||||
<p>创建优化器</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">99</span> <span class="bp">self</span><span class="o">.</span><span class="n">optimizer</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">optim</span><span class="o">.</span><span class="n">Adam</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">eps_model</span><span class="o">.</span><span class="n">parameters</span><span class="p">(),</span> <span class="n">lr</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">learning_rate</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-23'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-23'>#</a>
|
||||
</div>
|
||||
<p>图像日志记录</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">102</span> <span class="n">tracker</span><span class="o">.</span><span class="n">set_image</span><span class="p">(</span><span class="s2">"sample"</span><span class="p">,</span> <span class="kc">True</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-24'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-24'>#</a>
|
||||
</div>
|
||||
<h3>样本图片</h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">104</span> <span class="k">def</span> <span class="nf">sample</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-25'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-25'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">108</span> <span class="k">with</span> <span class="n">torch</span><span class="o">.</span><span class="n">no_grad</span><span class="p">():</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-26'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-26'>#</a>
|
||||
</div>
|
||||
<p><span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.58056em;vertical-align:-0.15em;"></span><span class="mord"><span class="mord mathnormal">x</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.32833099999999993em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight"><span class="mord mtight coloredeq eqf" style=""><span class="mord mathnormal mtight" style="margin-right:0.13889em">T</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span><span class="mspace" style="margin-right:0.2777777777777778em;"></span><span class="mrel">∼</span><span class="mspace" style="margin-right:0.2777777777777778em;"></span></span><span class="base"><span class="strut" style="height:1em;vertical-align:-0.25em;"></span><span class="mord mathnormal">p</span><span class="mopen">(</span><span class="mord"><span class="mord mathnormal">x</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.32833099999999993em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight"><span class="mord mtight coloredeq eqf" style=""><span class="mord mathnormal mtight" style="margin-right:0.13889em">T</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span><span class="mclose">)</span><span class="mspace" style="margin-right:0.2777777777777778em;"></span><span class="mrel">=</span><span class="mspace" style="margin-right:0.2777777777777778em;"></span></span><span class="base"><span class="strut" style="height:1em;vertical-align:-0.25em;"></span><span class="mord mathcal" style="margin-right:0.14736em;">N</span><span class="mopen">(</span><span class="mord"><span class="mord mathnormal">x</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.32833099999999993em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight"><span class="mord mtight coloredeq eqf" style=""><span class="mord mathnormal mtight" style="margin-right:0.13889em">T</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span><span class="mpunct">;</span><span class="mspace" style="margin-right:0.16666666666666666em;"></span><span class="mord mathbf">0</span><span class="mpunct">,</span><span class="mspace" style="margin-right:0.16666666666666666em;"></span><span class="mord mathbf">I</span><span class="mclose">)</span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">110</span> <span class="n">x</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">randn</span><span class="p">([</span><span class="bp">self</span><span class="o">.</span><span class="n">n_samples</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">image_channels</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">image_size</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">image_size</span><span class="p">],</span>
|
||||
<span class="lineno">111</span> <span class="n">device</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">device</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-27'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-27'>#</a>
|
||||
</div>
|
||||
<p>消除<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.68333em;vertical-align:0em;"></span><span class="mord coloredeq eqf" style=""><span class="mord mathnormal" style="margin-right:0.13889em">T</span></span></span></span></span></span>台阶噪音</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">114</span> <span class="k">for</span> <span class="n">t_</span> <span class="ow">in</span> <span class="n">monit</span><span class="o">.</span><span class="n">iterate</span><span class="p">(</span><span class="s1">'Sample'</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">n_steps</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-28'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-28'>#</a>
|
||||
</div>
|
||||
<p><span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.61508em;vertical-align:0em;"></span><span class="mord coloredeq eqg" style=""><span class="mord mathnormal" style="">t</span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">116</span> <span class="n">t</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">n_steps</span> <span class="o">-</span> <span class="n">t_</span> <span class="o">-</span> <span class="mi">1</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-29'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-29'>#</a>
|
||||
</div>
|
||||
<p>样本来自<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:1em;vertical-align:-0.25em;"></span><span class="mord" style="color:lightgreen"><span class="mord mathnormal" style="">p</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.33610799999999996em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mathnormal mtight" style="margin-right:0.02778em">θ</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span><span class="mopen">(</span><span class="mord"><span class="mord mathnormal">x</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.301108em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight"><span class="mord mtight"><span class="mord mtight coloredeq eqg" style=""><span class="mord mathnormal mtight" style="">t</span></span><span class="mbin mtight">−</span><span class="mord mtight">1</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.208331em;"><span></span></span></span></span></span></span><span class="mord">∣</span><span class="mord"><span class="mord mathnormal">x</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.2805559999999999em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight"><span class="mord mtight coloredeq eqg" style=""><span class="mord mathnormal mtight" style="">t</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span><span class="mclose">)</span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">118</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">diffusion</span><span class="o">.</span><span class="n">p_sample</span><span class="p">(</span><span class="n">x</span><span class="p">,</span> <span class="n">x</span><span class="o">.</span><span class="n">new_full</span><span class="p">((</span><span class="bp">self</span><span class="o">.</span><span class="n">n_samples</span><span class="p">,),</span> <span class="n">t</span><span class="p">,</span> <span class="n">dtype</span><span class="o">=</span><span class="n">torch</span><span class="o">.</span><span class="n">long</span><span class="p">))</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-30'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-30'>#</a>
|
||||
</div>
|
||||
<p>日志样本</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">121</span> <span class="n">tracker</span><span class="o">.</span><span class="n">save</span><span class="p">(</span><span class="s1">'sample'</span><span class="p">,</span> <span class="n">x</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-31'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-31'>#</a>
|
||||
</div>
|
||||
<h3>火车</h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">123</span> <span class="k">def</span> <span class="nf">train</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-32'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-32'>#</a>
|
||||
</div>
|
||||
<p>遍历数据集</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">129</span> <span class="k">for</span> <span class="n">data</span> <span class="ow">in</span> <span class="n">monit</span><span class="o">.</span><span class="n">iterate</span><span class="p">(</span><span class="s1">'Train'</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">data_loader</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-33'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-33'>#</a>
|
||||
</div>
|
||||
<p>递增全局步长</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">131</span> <span class="n">tracker</span><span class="o">.</span><span class="n">add_global_step</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-34'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-34'>#</a>
|
||||
</div>
|
||||
<p>将数据移动到设备</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">133</span> <span class="n">data</span> <span class="o">=</span> <span class="n">data</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">device</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-35'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-35'>#</a>
|
||||
</div>
|
||||
<p>将渐变设为零</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">136</span> <span class="bp">self</span><span class="o">.</span><span class="n">optimizer</span><span class="o">.</span><span class="n">zero_grad</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-36'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-36'>#</a>
|
||||
</div>
|
||||
<p>计算损失</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">138</span> <span class="n">loss</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">diffusion</span><span class="o">.</span><span class="n">loss</span><span class="p">(</span><span class="n">data</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-37'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-37'>#</a>
|
||||
</div>
|
||||
<p>计算梯度</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">140</span> <span class="n">loss</span><span class="o">.</span><span class="n">backward</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-38'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-38'>#</a>
|
||||
</div>
|
||||
<p>采取优化步骤</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">142</span> <span class="bp">self</span><span class="o">.</span><span class="n">optimizer</span><span class="o">.</span><span class="n">step</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-39'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-39'>#</a>
|
||||
</div>
|
||||
<p>追踪损失</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">144</span> <span class="n">tracker</span><span class="o">.</span><span class="n">save</span><span class="p">(</span><span class="s1">'loss'</span><span class="p">,</span> <span class="n">loss</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-40'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-40'>#</a>
|
||||
</div>
|
||||
<h3>训练循环</h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">146</span> <span class="k">def</span> <span class="nf">run</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-41'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-41'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">150</span> <span class="k">for</span> <span class="n">_</span> <span class="ow">in</span> <span class="n">monit</span><span class="o">.</span><span class="n">loop</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">epochs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-42'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-42'>#</a>
|
||||
</div>
|
||||
<p>训练模型</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">152</span> <span class="bp">self</span><span class="o">.</span><span class="n">train</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-43'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-43'>#</a>
|
||||
</div>
|
||||
<p>对一些图像进行采样</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">154</span> <span class="bp">self</span><span class="o">.</span><span class="n">sample</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-44'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-44'>#</a>
|
||||
</div>
|
||||
<p>控制台中的新行</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">156</span> <span class="n">tracker</span><span class="o">.</span><span class="n">new_line</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-45'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-45'>#</a>
|
||||
</div>
|
||||
<p>保存模型</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">158</span> <span class="n">experiment</span><span class="o">.</span><span class="n">save_checkpoint</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-46'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-46'>#</a>
|
||||
</div>
|
||||
<h3>CeleBA HQ 数据集</h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">161</span><span class="k">class</span> <span class="nc">CelebADataset</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">utils</span><span class="o">.</span><span class="n">data</span><span class="o">.</span><span class="n">Dataset</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-47'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-47'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">166</span> <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">image_size</span><span class="p">:</span> <span class="nb">int</span><span class="p">):</span>
|
||||
<span class="lineno">167</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-48'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-48'>#</a>
|
||||
</div>
|
||||
<p>CeleBA 图片文件夹</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">170</span> <span class="n">folder</span> <span class="o">=</span> <span class="n">lab</span><span class="o">.</span><span class="n">get_data_path</span><span class="p">()</span> <span class="o">/</span> <span class="s1">'celebA'</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-49'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-49'>#</a>
|
||||
</div>
|
||||
<p>文件清单</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">172</span> <span class="bp">self</span><span class="o">.</span><span class="n">_files</span> <span class="o">=</span> <span class="p">[</span><span class="n">p</span> <span class="k">for</span> <span class="n">p</span> <span class="ow">in</span> <span class="n">folder</span><span class="o">.</span><span class="n">glob</span><span class="p">(</span><span class="sa">f</span><span class="s1">'**/*.jpg'</span><span class="p">)]</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-50'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-50'>#</a>
|
||||
</div>
|
||||
<p>用于调整图像大小并转换为张量的转换</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">175</span> <span class="bp">self</span><span class="o">.</span><span class="n">_transform</span> <span class="o">=</span> <span class="n">torchvision</span><span class="o">.</span><span class="n">transforms</span><span class="o">.</span><span class="n">Compose</span><span class="p">([</span>
|
||||
<span class="lineno">176</span> <span class="n">torchvision</span><span class="o">.</span><span class="n">transforms</span><span class="o">.</span><span class="n">Resize</span><span class="p">(</span><span class="n">image_size</span><span class="p">),</span>
|
||||
<span class="lineno">177</span> <span class="n">torchvision</span><span class="o">.</span><span class="n">transforms</span><span class="o">.</span><span class="n">ToTensor</span><span class="p">(),</span>
|
||||
<span class="lineno">178</span> <span class="p">])</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-51'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-51'>#</a>
|
||||
</div>
|
||||
<p>数据集的大小</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">180</span> <span class="k">def</span> <span class="fm">__len__</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-52'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-52'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">184</span> <span class="k">return</span> <span class="nb">len</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">_files</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-53'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-53'>#</a>
|
||||
</div>
|
||||
<p>获取一张图片</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">186</span> <span class="k">def</span> <span class="fm">__getitem__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">index</span><span class="p">:</span> <span class="nb">int</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-54'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-54'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">190</span> <span class="n">img</span> <span class="o">=</span> <span class="n">Image</span><span class="o">.</span><span class="n">open</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">_files</span><span class="p">[</span><span class="n">index</span><span class="p">])</span>
|
||||
<span class="lineno">191</span> <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">_transform</span><span class="p">(</span><span class="n">img</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-55'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-55'>#</a>
|
||||
</div>
|
||||
<p>创建 CeleBA 数据集</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">194</span><span class="nd">@option</span><span class="p">(</span><span class="n">Configs</span><span class="o">.</span><span class="n">dataset</span><span class="p">,</span> <span class="s1">'CelebA'</span><span class="p">)</span>
|
||||
<span class="lineno">195</span><span class="k">def</span> <span class="nf">celeb_dataset</span><span class="p">(</span><span class="n">c</span><span class="p">:</span> <span class="n">Configs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-56'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-56'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">199</span> <span class="k">return</span> <span class="n">CelebADataset</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">image_size</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-57'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-57'>#</a>
|
||||
</div>
|
||||
<h3>MNIST 数据集</h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">202</span><span class="k">class</span> <span class="nc">MNISTDataset</span><span class="p">(</span><span class="n">torchvision</span><span class="o">.</span><span class="n">datasets</span><span class="o">.</span><span class="n">MNIST</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-58'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-58'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">207</span> <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">image_size</span><span class="p">):</span>
|
||||
<span class="lineno">208</span> <span class="n">transform</span> <span class="o">=</span> <span class="n">torchvision</span><span class="o">.</span><span class="n">transforms</span><span class="o">.</span><span class="n">Compose</span><span class="p">([</span>
|
||||
<span class="lineno">209</span> <span class="n">torchvision</span><span class="o">.</span><span class="n">transforms</span><span class="o">.</span><span class="n">Resize</span><span class="p">(</span><span class="n">image_size</span><span class="p">),</span>
|
||||
<span class="lineno">210</span> <span class="n">torchvision</span><span class="o">.</span><span class="n">transforms</span><span class="o">.</span><span class="n">ToTensor</span><span class="p">(),</span>
|
||||
<span class="lineno">211</span> <span class="p">])</span>
|
||||
<span class="lineno">212</span>
|
||||
<span class="lineno">213</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">(</span><span class="nb">str</span><span class="p">(</span><span class="n">lab</span><span class="o">.</span><span class="n">get_data_path</span><span class="p">()),</span> <span class="n">train</span><span class="o">=</span><span class="kc">True</span><span class="p">,</span> <span class="n">download</span><span class="o">=</span><span class="kc">True</span><span class="p">,</span> <span class="n">transform</span><span class="o">=</span><span class="n">transform</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-59'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-59'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">215</span> <span class="k">def</span> <span class="fm">__getitem__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">item</span><span class="p">):</span>
|
||||
<span class="lineno">216</span> <span class="k">return</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__getitem__</span><span class="p">(</span><span class="n">item</span><span class="p">)[</span><span class="mi">0</span><span class="p">]</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-60'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-60'>#</a>
|
||||
</div>
|
||||
<p>创建 MNIST 数据集</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">219</span><span class="nd">@option</span><span class="p">(</span><span class="n">Configs</span><span class="o">.</span><span class="n">dataset</span><span class="p">,</span> <span class="s1">'MNIST'</span><span class="p">)</span>
|
||||
<span class="lineno">220</span><span class="k">def</span> <span class="nf">mnist_dataset</span><span class="p">(</span><span class="n">c</span><span class="p">:</span> <span class="n">Configs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-61'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-61'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">224</span> <span class="k">return</span> <span class="n">MNISTDataset</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">image_size</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-62'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-62'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">227</span><span class="k">def</span> <span class="nf">main</span><span class="p">():</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-63'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-63'>#</a>
|
||||
</div>
|
||||
<p>创建实验</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">229</span> <span class="n">experiment</span><span class="o">.</span><span class="n">create</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">'diffuse'</span><span class="p">,</span> <span class="n">writers</span><span class="o">=</span><span class="p">{</span><span class="s1">'screen'</span><span class="p">,</span> <span class="s1">'labml'</span><span class="p">})</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-64'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-64'>#</a>
|
||||
</div>
|
||||
<p>创建配置</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">232</span> <span class="n">configs</span> <span class="o">=</span> <span class="n">Configs</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-65'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-65'>#</a>
|
||||
</div>
|
||||
<p>设置配置。您可以通过在字典中传递值来覆盖默认值。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">235</span> <span class="n">experiment</span><span class="o">.</span><span class="n">configs</span><span class="p">(</span><span class="n">configs</span><span class="p">,</span> <span class="p">{</span>
|
||||
<span class="lineno">236</span> <span class="s1">'dataset'</span><span class="p">:</span> <span class="s1">'CelebA'</span><span class="p">,</span> <span class="c1"># 'MNIST'</span>
|
||||
<span class="lineno">237</span> <span class="s1">'image_channels'</span><span class="p">:</span> <span class="mi">3</span><span class="p">,</span> <span class="c1"># 1,</span>
|
||||
<span class="lineno">238</span> <span class="s1">'epochs'</span><span class="p">:</span> <span class="mi">100</span><span class="p">,</span> <span class="c1"># 5,</span>
|
||||
<span class="lineno">239</span> <span class="p">})</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-66'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-66'>#</a>
|
||||
</div>
|
||||
<p>初始化</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">242</span> <span class="n">configs</span><span class="o">.</span><span class="n">init</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-67'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-67'>#</a>
|
||||
</div>
|
||||
<p>设置用于保存和加载的模型</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">245</span> <span class="n">experiment</span><span class="o">.</span><span class="n">add_pytorch_models</span><span class="p">({</span><span class="s1">'eps_model'</span><span class="p">:</span> <span class="n">configs</span><span class="o">.</span><span class="n">eps_model</span><span class="p">})</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-68'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-68'>#</a>
|
||||
</div>
|
||||
<p>启动并运行训练循环</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">248</span> <span class="k">with</span> <span class="n">experiment</span><span class="o">.</span><span class="n">start</span><span class="p">():</span>
|
||||
<span class="lineno">249</span> <span class="n">configs</span><span class="o">.</span><span class="n">run</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-69'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-69'>#</a>
|
||||
</div>
|
||||
<p></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">253</span><span class="k">if</span> <span class="vm">__name__</span> <span class="o">==</span> <span class="s1">'__main__'</span><span class="p">:</span>
|
||||
<span class="lineno">254</span> <span class="n">main</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
File diff suppressed because one or more lines are too long
@@ -0,0 +1,130 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content=""/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="去噪扩散概率模型 (DDPM)"/>
|
||||
<meta name="twitter:description" content=""/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/diffusion/ddpm/readme.html"/>
|
||||
<meta property="og:title" content="去噪扩散概率模型 (DDPM)"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="去噪扩散概率模型 (DDPM)"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="去噪扩散概率模型 (DDPM)"/>
|
||||
<meta property="og:description" content=""/>
|
||||
|
||||
<title>去噪扩散概率模型 (DDPM)</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/diffusion/ddpm/readme.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="../index.html">diffusion</a>
|
||||
<a class="parent" href="index.html">ddpm</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/diffusion/ddpm/readme.md" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1><a href="https://nn.labml.ai/diffusion/ddpm/index.html">去噪扩散概率模型 (DDPM)</a></h1>
|
||||
<p><a href="https://colab.research.google.com/github/labmlai/annotated_deep_learning_paper_implementations/blob/master/labml_nn/diffusion/ddpm/experiment.ipynb"><img alt="Open In Colab" src="https://colab.research.google.com/assets/colab-badge.svg"></a></p>
|
||||
<p>这是《<a href="https://arxiv.org/abs/2006.11239">去噪扩散概率模型</a>》论文的 <a href="https://pytorch.org">PyTorch</a> 实现/教程。</p>
|
||||
<p>简而言之，我们从数据中获取图像并逐步添加噪点。然后，我们训练一个模型来预测每个步骤的噪声，并使用该模型生成图像。</p>
|
||||
<p>这是预测噪声和<a href="https://nn.labml.ai/diffusion/ddpm/experiment.html">训练代码</a>的 <a href="https://nn.labml.ai/diffusion/ddpm/unet.html">UNet 模型</a>。<a href="https://nn.labml.ai/diffusion/ddpm/evaluate.html">此文件</a>可以从经过训练的模型生成样本和插值。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
File diff suppressed because one or more lines are too long
@@ -0,0 +1,150 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content="DDPM 实验的实用程序函数"/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="DDPM 实验的实用程序函数"/>
|
||||
<meta name="twitter:description" content="DDPM 实验的实用程序函数"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/diffusion/ddpm/utils.html"/>
|
||||
<meta property="og:title" content="DDPM 实验的实用程序函数"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="DDPM 实验的实用程序函数"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="DDPM 实验的实用程序函数"/>
|
||||
<meta property="og:description" content="DDPM 实验的实用程序函数"/>
|
||||
|
||||
<title>DDPM 实验的实用程序函数</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/diffusion/ddpm/utils.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="../index.html">diffusion</a>
|
||||
<a class="parent" href="index.html">ddpm</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/diffusion/ddpm/utils.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1><a href="index.html">DDPM</a> 实验的实用程序函数</h1>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">10</span><span></span><span class="kn">import</span> <span class="nn">torch.utils.data</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-1'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-1'>#</a>
|
||||
</div>
|
||||
<p>收集要素地图形状的<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.61508em;vertical-align:0em;"></span><span class="mord mathnormal">t</span></span></span></span></span>常量并将其整形为要素地图形状</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">13</span><span class="k">def</span> <span class="nf">gather</span><span class="p">(</span><span class="n">consts</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span> <span class="n">t</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-2'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-2'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">15</span> <span class="n">c</span> <span class="o">=</span> <span class="n">consts</span><span class="o">.</span><span class="n">gather</span><span class="p">(</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="n">t</span><span class="p">)</span>
|
||||
<span class="lineno">16</span> <span class="k">return</span> <span class="n">c</span><span class="o">.</span><span class="n">reshape</span><span class="p">(</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
@@ -0,0 +1,129 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content="一组关于扩散模型的 PyTorch 实现/教程。"/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="扩散模型"/>
|
||||
<meta name="twitter:description" content="一组关于扩散模型的 PyTorch 实现/教程。"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/diffusion/index.html"/>
|
||||
<meta property="og:title" content="扩散模型"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="扩散模型"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="扩散模型"/>
|
||||
<meta property="og:description" content="一组关于扩散模型的 PyTorch 实现/教程。"/>
|
||||
|
||||
<title>扩散模型</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/diffusion/index.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="index.html">diffusion</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/diffusion/__init__.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1>扩散模型</h1>
|
||||
<ul><li><a href="ddpm/index.html">去噪扩散概率模型 (DDPM)</a></li>
|
||||
<li><a href="stable_diffusion/index.html">稳定的扩散</a></li>
|
||||
<li><a href="stable_diffusion/latent_diffusion.html">潜在扩散模型</a></li>
|
||||
<li><a href="stable_diffusion/sampler/ddim.html">降噪扩散隐含模型 (DDIM) 采样</a></li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
@@ -0,0 +1,148 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content="带注释的 PyTorch 稳定扩散实现/教程。"/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="稳定的扩散"/>
|
||||
<meta name="twitter:description" content="带注释的 PyTorch 稳定扩散实现/教程。"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/diffusion/stable_diffusion/index.html"/>
|
||||
<meta property="og:title" content="稳定的扩散"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="稳定的扩散"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="稳定的扩散"/>
|
||||
<meta property="og:description" content="带注释的 PyTorch 稳定扩散实现/教程。"/>
|
||||
|
||||
<title>稳定的扩散</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/diffusion/stable_diffusion/index.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="../index.html">diffusion</a>
|
||||
<a class="parent" href="index.html">stable_diffusion</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/diffusion/stable_diffusion/__init__.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1>稳定的扩散</h1>
|
||||
<p>这是基于官方的稳定扩散存储库 <a href="https://github.com/CompVis/stable-diffusion">compVIS/Stable-</a> Difusion。我们保持了模型结构不变，以便可以直接加载开源权重。我们的实现不包含训练代码。</p>
|
||||
<h3><a href="https://promptart.labml.ai">PrompA</a></h3>
|
||||
<p><img alt="PromptArt" src="https://labml.ai/images/promptart-feed.webp"></p>
|
||||
<p>我们已经在 p <a href="https://promptart.labml.ai">romptart.labml.ai</a> 部署了基于扩散的稳定图像生成服务</p>
|
||||
<h3><a href="latent_diffusion.html">潜在扩散模型</a></h3>
|
||||
<p>核心是<a href="latent_diffusion.html">潜在扩散模型</a>。它包括：</p>
|
||||
<ul><li><a href="model/autoencoder.html">自动编码器</a></li>
|
||||
<li><a href="model/unet_attention.html">备受关注</a>的 <a href="model/unet.html">U-Net</a></li></ul>
|
||||
<p>我们还（可选）将 <a href="https://github.com/HazyResearch/flash-attention">Flash</a> Actontion 集成到我们的 <a href="model/unet_attention.html">U-Net 注意力</a>中，这使您可以在 RTX A6000 GPU 上将性能提高近50％。</p>
|
||||
<p>扩散是基于 <a href="model/clip_embedder.html">CLIP 嵌入</a>来调节的。</p>
|
||||
<h3><a href="sampler/index.html">采样算法</a></h3>
|
||||
<p>我们已经实现了以下<a href="sampler/index.html">采样算法</a>：</p>
|
||||
<ul><li><a href="sampler/ddpm.html">去噪扩散概率模型 (DDPM) 采样</a></li>
|
||||
<li><a href="sampler/ddim.html">降噪扩散隐含模型 (DDIM) 采样</a></li></ul>
|
||||
<h3><a href="scripts/index.html">示例脚本</a></h3>
|
||||
<p>以下是图像生成脚本：</p>
|
||||
<li><a href="scripts/text_to_image.html">根据文字提示生成图像</a></li> <ul>
|
||||
<li><a href="scripts/image_to_image.html">在提示的指导下，根据给定的图像生成图像</a></li>
|
||||
<li><a href="scripts/in_paint.html">根据文本提示修改给定图像的各个部分</a></li></ul>
|
||||
<h4><a href="util.html">公共事业</a></h4>
|
||||
<p><a href="util.html"><code class="highlight"><span></span><span class="n">util</span><span class="o">.</span><span class="n">py</span></code>
|
||||
</a>定义了实用函数。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
@@ -0,0 +1,453 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content="带注释的 PyTorch 实现/教程来自论文的潜在扩散模型使用潜在扩散模型进行高分辨率图像合成"/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="潜在扩散模型"/>
|
||||
<meta name="twitter:description" content="带注释的 PyTorch 实现/教程来自论文的潜在扩散模型使用潜在扩散模型进行高分辨率图像合成"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/diffusion/stable_diffusion/latent_diffusion.html"/>
|
||||
<meta property="og:title" content="潜在扩散模型"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="潜在扩散模型"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="潜在扩散模型"/>
|
||||
<meta property="og:description" content="带注释的 PyTorch 实现/教程来自论文的潜在扩散模型使用潜在扩散模型进行高分辨率图像合成"/>
|
||||
|
||||
<title>潜在扩散模型</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/diffusion/stable_diffusion/latent_diffusion.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="../index.html">diffusion</a>
|
||||
<a class="parent" href="index.html">stable_diffusion</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/diffusion/stable_diffusion/latent_diffusion.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1>潜在扩散模型</h1>
|
||||
<p>潜在扩散模型使用自动编码器在图像空间和潜在空间之间进行映射。扩散模型适用于潜在空间，这使得训练变得容易得多。它基于<a href="https://arxiv.org/abs/2112.10752">带有潜在扩散模型的纸质高分辨率图像合成</a>。</p>
|
||||
<p>它们使用预训练的自动编码器，在预训练的自动编码器的潜在空间上训练扩散 U-Net。</p>
|
||||
<p>有关更简单的扩散实现，请参阅我们的 <a href="../ddpm/index.html">DDPM 实现</a>。我们对<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.58056em;vertical-align:-0.15em;"></span><span class="mord coloredeq eqd" style=""><span class="mord" style=""><span class="mord mathnormal" style="margin-right:0.0037em">α</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.2805559999999999em;"><span style="top:-2.5500000000000003em;margin-left:-0.0037em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style=""><span class="mord mathnormal mtight coloredeq eqj" style="">t</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span></span></span></span></span></span><span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.8888799999999999em;vertical-align:-0.19444em;"></span><span class="mord coloredeq eqe" style=""><span class="mord" style=""><span class="mord" style=""><span class="mord mathnormal coloredeq eqf" style="margin-right:0.05278em">β</span></span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.2805559999999999em;"><span style="top:-2.5500000000000003em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style=""><span class="mord mathnormal mtight coloredeq eqj" style="">t</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span></span></span></span></span></span>时间表等使用相同的符号。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">24</span><span></span><span class="kn">from</span> <span class="nn">typing</span> <span class="kn">import</span> <span class="n">List</span>
|
||||
<span class="lineno">25</span>
|
||||
<span class="lineno">26</span><span class="kn">import</span> <span class="nn">torch</span>
|
||||
<span class="lineno">27</span><span class="kn">import</span> <span class="nn">torch.nn</span> <span class="k">as</span> <span class="nn">nn</span>
|
||||
<span class="lineno">28</span>
|
||||
<span class="lineno">29</span><span class="kn">from</span> <span class="nn">labml_nn.diffusion.stable_diffusion.model.autoencoder</span> <span class="kn">import</span> <span class="n">Autoencoder</span>
|
||||
<span class="lineno">30</span><span class="kn">from</span> <span class="nn">labml_nn.diffusion.stable_diffusion.model.clip_embedder</span> <span class="kn">import</span> <span class="n">CLIPTextEmbedder</span>
|
||||
<span class="lineno">31</span><span class="kn">from</span> <span class="nn">labml_nn.diffusion.stable_diffusion.model.unet</span> <span class="kn">import</span> <span class="n">UNetModel</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-1'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-1'>#</a>
|
||||
</div>
|
||||
<p><em>这是围绕 <a href="model/unet.html">U-Net</a> 的空包装类。我们保持它与 <a href="https://github.com/CompVis/stable-diffusion">compVIS/Stable-</a> Difusion 相同的模型结构，这样我们就不必明确地映射检查点权重</em>。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">34</span><span class="k">class</span> <span class="nc">DiffusionWrapper</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-2'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-2'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">42</span> <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">diffusion_model</span><span class="p">:</span> <span class="n">UNetModel</span><span class="p">):</span>
|
||||
<span class="lineno">43</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
|
||||
<span class="lineno">44</span> <span class="bp">self</span><span class="o">.</span><span class="n">diffusion_model</span> <span class="o">=</span> <span class="n">diffusion_model</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-3'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-3'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">46</span> <span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span> <span class="n">time_steps</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span> <span class="n">context</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">):</span>
|
||||
<span class="lineno">47</span> <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">diffusion_model</span><span class="p">(</span><span class="n">x</span><span class="p">,</span> <span class="n">time_steps</span><span class="p">,</span> <span class="n">context</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-4'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-4'>#</a>
|
||||
</div>
|
||||
<h2>潜在扩散模型</h2>
|
||||
<p>它包含以下组件：</p>
|
||||
<ul><li><a href="model/autoencoder.html">自动编码器</a></li>
|
||||
<li><a href="model/unet_attention.html">备受关注</a>的 <a href="model/unet.html">U-Net</a></li>
|
||||
<li><a href="model/clip_embedder.html">CLIP 嵌入式生成器</a></li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">50</span><span class="k">class</span> <span class="nc">LatentDiffusion</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-5'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-5'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">60</span> <span class="n">model</span><span class="p">:</span> <span class="n">DiffusionWrapper</span>
|
||||
<span class="lineno">61</span> <span class="n">first_stage_model</span><span class="p">:</span> <span class="n">Autoencoder</span>
|
||||
<span class="lineno">62</span> <span class="n">cond_stage_model</span><span class="p">:</span> <span class="n">CLIPTextEmbedder</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-6'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-6'>#</a>
|
||||
</div>
|
||||
<ul><li><code class="highlight"><span></span><span class="n">unet_model</span></code>
|
||||
是预测潜在空间中噪声<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:1em;vertical-align:-0.25em;"></span><span class="mord coloredeq eqb" style=""><span class="mord" style=""><span class="mord mathnormal" style="">ϵ</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.33610799999999996em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord text mtight" style=""><span class="mord mtight" style=""><span class="mord mtight coloredeq eqi" style="">c</span></span><span class="mord mtight" style="">ond</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span><span class="mopen" style="">(</span><span class="mord" style=""><span class="mord coloredeq eqg" style=""><span class="mord mathnormal" style="">x</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.2805559999999999em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style=""><span class="mord mathnormal mtight coloredeq eqj" style="">t</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span></span><span class="mpunct" style="">,</span><span class="mspace" style="margin-right:0.16666666666666666em;"></span><span class="mord" style=""><span class="mord mathnormal coloredeq eqi" style="">c</span></span><span class="mclose" style="">)</span></span></span></span></span></span>的 <a href="model/unet.html">U-Ne</a> t</li>
|
||||
<li><code class="highlight"><span></span><span class="n">autoencoder</span></code>
|
||||
是<a href="model/autoencoder.html">自动编码器</a></li>
|
||||
<li><code class="highlight"><span></span><span class="n">clip_embedder</span></code>
|
||||
是 <a href="model/clip_embedder.html">CLIP 嵌入生成器</a></li>
|
||||
<li><code class="highlight"><span></span><span class="n">latent_scaling_factor</span></code>
|
||||
是潜在空间的缩放系数。在馈入 U-Net 之前，自动编码器的编码会按此进行缩放。</li>
|
||||
<li><code class="highlight"><span></span><span class="n">n_steps</span></code>
|
||||
是扩散步骤的数量<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.68333em;vertical-align:0em;"></span><span class="mord coloredeq eqh" style=""><span class="mord mathnormal" style="margin-right:0.13889em">T</span></span></span></span></span></span>。</li>
|
||||
<li><code class="highlight"><span></span><span class="n">linear_start</span></code>
|
||||
是<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.8888799999999999em;vertical-align:-0.19444em;"></span><span class="mord coloredeq eqf" style=""><span class="mord mathnormal" style="margin-right:0.05278em">β</span></span></span></span></span></span>时间表的开始。</li>
|
||||
<li><code class="highlight"><span></span><span class="n">linear_end</span></code>
|
||||
是<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.8888799999999999em;vertical-align:-0.19444em;"></span><span class="mord coloredeq eqf" style=""><span class="mord mathnormal" style="margin-right:0.05278em">β</span></span></span></span></span></span>时间表的结束。</li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">64</span> <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span>
|
||||
<span class="lineno">65</span> <span class="n">unet_model</span><span class="p">:</span> <span class="n">UNetModel</span><span class="p">,</span>
|
||||
<span class="lineno">66</span> <span class="n">autoencoder</span><span class="p">:</span> <span class="n">Autoencoder</span><span class="p">,</span>
|
||||
<span class="lineno">67</span> <span class="n">clip_embedder</span><span class="p">:</span> <span class="n">CLIPTextEmbedder</span><span class="p">,</span>
|
||||
<span class="lineno">68</span> <span class="n">latent_scaling_factor</span><span class="p">:</span> <span class="nb">float</span><span class="p">,</span>
|
||||
<span class="lineno">69</span> <span class="n">n_steps</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span>
|
||||
<span class="lineno">70</span> <span class="n">linear_start</span><span class="p">:</span> <span class="nb">float</span><span class="p">,</span>
|
||||
<span class="lineno">71</span> <span class="n">linear_end</span><span class="p">:</span> <span class="nb">float</span><span class="p">,</span>
|
||||
<span class="lineno">72</span> <span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-7'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-7'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">84</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-8'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-8'>#</a>
|
||||
</div>
|
||||
<p>封装 <a href="model/unet.html">U-Net</a> 以保持与 <a href="https://github.com/CompVis/stable-diffusion">compVIS/Stable-</a> Difusion 相同的模型结构。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">87</span> <span class="bp">self</span><span class="o">.</span><span class="n">model</span> <span class="o">=</span> <span class="n">DiffusionWrapper</span><span class="p">(</span><span class="n">unet_model</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-9'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-9'>#</a>
|
||||
</div>
|
||||
<p>自动编码器和缩放系数</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">89</span> <span class="bp">self</span><span class="o">.</span><span class="n">first_stage_model</span> <span class="o">=</span> <span class="n">autoencoder</span>
|
||||
<span class="lineno">90</span> <span class="bp">self</span><span class="o">.</span><span class="n">latent_scaling_factor</span> <span class="o">=</span> <span class="n">latent_scaling_factor</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-10'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-10'>#</a>
|
||||
</div>
|
||||
<p><a href="model/clip_embedder.html">CLIP 嵌入式生成器</a></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">92</span> <span class="bp">self</span><span class="o">.</span><span class="n">cond_stage_model</span> <span class="o">=</span> <span class="n">clip_embedder</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-11'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-11'>#</a>
|
||||
</div>
|
||||
<p>步数<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.68333em;vertical-align:0em;"></span><span class="mord coloredeq eqh" style=""><span class="mord mathnormal" style="margin-right:0.13889em">T</span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">95</span> <span class="bp">self</span><span class="o">.</span><span class="n">n_steps</span> <span class="o">=</span> <span class="n">n_steps</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-12'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-12'>#</a>
|
||||
</div>
|
||||
<p><span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.8888799999999999em;vertical-align:-0.19444em;"></span><span class="mord coloredeq eqf" style=""><span class="mord mathnormal" style="margin-right:0.05278em">β</span></span></span></span></span></span>时间表</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">98</span> <span class="n">beta</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">linspace</span><span class="p">(</span><span class="n">linear_start</span> <span class="o">**</span> <span class="mf">0.5</span><span class="p">,</span> <span class="n">linear_end</span> <span class="o">**</span> <span class="mf">0.5</span><span class="p">,</span> <span class="n">n_steps</span><span class="p">,</span> <span class="n">dtype</span><span class="o">=</span><span class="n">torch</span><span class="o">.</span><span class="n">float64</span><span class="p">)</span> <span class="o">**</span> <span class="mi">2</span>
|
||||
<span class="lineno">99</span> <span class="bp">self</span><span class="o">.</span><span class="n">beta</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Parameter</span><span class="p">(</span><span class="n">beta</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">float32</span><span class="p">),</span> <span class="n">requires_grad</span><span class="o">=</span><span class="kc">False</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-13'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-13'>#</a>
|
||||
</div>
|
||||
<p><span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.58056em;vertical-align:-0.15em;"></span><span class="mord coloredeq eqd" style=""><span class="mord" style=""><span class="mord mathnormal" style="margin-right:0.0037em">α</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.2805559999999999em;"><span style="top:-2.5500000000000003em;margin-left:-0.0037em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style=""><span class="mord mathnormal mtight coloredeq eqj" style="">t</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span></span><span class="mspace" style="margin-right:0.2777777777777778em;"></span><span class="mrel">=</span><span class="mspace" style="margin-right:0.2777777777777778em;"></span></span><span class="base"><span class="strut" style="height:0.72777em;vertical-align:-0.08333em;"></span><span class="mord">1</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin">−</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span></span><span class="base"><span class="strut" style="height:0.8888799999999999em;vertical-align:-0.19444em;"></span><span class="mord coloredeq eqe" style=""><span class="mord" style=""><span class="mord" style=""><span class="mord mathnormal coloredeq eqf" style="margin-right:0.05278em">β</span></span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.2805559999999999em;"><span style="top:-2.5500000000000003em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style=""><span class="mord mathnormal mtight coloredeq eqj" style="">t</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">101</span> <span class="n">alpha</span> <span class="o">=</span> <span class="mf">1.</span> <span class="o">-</span> <span class="n">beta</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-14'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-14'>#</a>
|
||||
</div>
|
||||
<p><span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.71778em;vertical-align:-0.15em;"></span><span class="mord accent"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.56778em;"><span style="top:-3em;"><span class="pstrut" style="height:3em;"></span><span class="mord coloredeq eqd" style=""><span class="mord mathnormal" style="margin-right:0.0037em">α</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.2805559999999999em;"><span style="top:-2.5500000000000003em;margin-left:-0.0037em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style=""><span class="mord mathnormal mtight coloredeq eqj" style="">t</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span></span><span style="top:-3em;"><span class="pstrut" style="height:3em;"></span><span class="accent-body" style="left:-0.25em;"><span class="mord">ˉ</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span><span class="mspace" style="margin-right:0.2777777777777778em;"></span><span class="mrel">=</span><span class="mspace" style="margin-right:0.2777777777777778em;"></span></span><span class="base"><span class="strut" style="height:1.233166em;vertical-align:-0.29971000000000003em;"></span><span class="mop"><span class="mop op-symbol small-op" style="position:relative;top:-0.0000050000000000050004em;">∏</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.933456em;"><span style="top:-2.40029em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight"><span class="mord mtight"><span class="mord mathnormal mtight">s</span><span class="mrel mtight">=</span><span class="mord mtight">1</span></span></span></span><span style="top:-3.2029em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight"><span class="mord mtight coloredeq eqj" style=""><span class="mord mathnormal mtight" style="">t</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.29971000000000003em;"><span></span></span></span></span></span></span><span class="mspace" style="margin-right:0.16666666666666666em;"></span><span class="mord"><span class="mord mathnormal" style="margin-right:0.0037em;">α</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.151392em;"><span style="top:-2.5500000000000003em;margin-left:-0.0037em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight"><span class="mord mathnormal mtight">s</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">103</span> <span class="n">alpha_bar</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">cumprod</span><span class="p">(</span><span class="n">alpha</span><span class="p">,</span> <span class="n">dim</span><span class="o">=</span><span class="mi">0</span><span class="p">)</span>
|
||||
<span class="lineno">104</span> <span class="bp">self</span><span class="o">.</span><span class="n">alpha_bar</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Parameter</span><span class="p">(</span><span class="n">alpha_bar</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">float32</span><span class="p">),</span> <span class="n">requires_grad</span><span class="o">=</span><span class="kc">False</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-15'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-15'>#</a>
|
||||
</div>
|
||||
<h3>获取设备模型</h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">106</span> <span class="nd">@property</span>
|
||||
<span class="lineno">107</span> <span class="k">def</span> <span class="nf">device</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-16'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-16'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">111</span> <span class="k">return</span> <span class="nb">next</span><span class="p">(</span><span class="nb">iter</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">model</span><span class="o">.</span><span class="n">parameters</span><span class="p">()))</span><span class="o">.</span><span class="n">device</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-17'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-17'>#</a>
|
||||
</div>
|
||||
<h3>获取 <a href="model/clip_embedder.html">CLIP 嵌入</a>以获取文本提示列表</h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">113</span> <span class="k">def</span> <span class="nf">get_text_conditioning</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">prompts</span><span class="p">:</span> <span class="n">List</span><span class="p">[</span><span class="nb">str</span><span class="p">]):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-18'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-18'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">117</span> <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">cond_stage_model</span><span class="p">(</span><span class="n">prompts</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-19'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-19'>#</a>
|
||||
</div>
|
||||
<h3>获取图像的缩放潜在空间表示</h3>
|
||||
<p>编码器输出是分布式。我们从中取样并乘以缩放系数。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">119</span> <span class="k">def</span> <span class="nf">autoencoder_encode</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">image</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-20'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-20'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">126</span> <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">latent_scaling_factor</span> <span class="o">*</span> <span class="bp">self</span><span class="o">.</span><span class="n">first_stage_model</span><span class="o">.</span><span class="n">encode</span><span class="p">(</span><span class="n">image</span><span class="p">)</span><span class="o">.</span><span class="n">sample</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-21'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-21'>#</a>
|
||||
</div>
|
||||
<h3>从潜在表示中获取图像</h3>
|
||||
<p>我们按缩放系数向下缩放，然后解码。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">128</span> <span class="k">def</span> <span class="nf">autoencoder_decode</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">z</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-22'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-22'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">134</span> <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">first_stage_model</span><span class="o">.</span><span class="n">decode</span><span class="p">(</span><span class="n">z</span> <span class="o">/</span> <span class="bp">self</span><span class="o">.</span><span class="n">latent_scaling_factor</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-23'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-23'>#</a>
|
||||
</div>
|
||||
<h3>预测噪音</h3>
|
||||
<p>根据潜在表示<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.58056em;vertical-align:-0.15em;"></span><span class="mord coloredeq eqg" style=""><span class="mord" style=""><span class="mord mathnormal" style="">x</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.2805559999999999em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style=""><span class="mord mathnormal mtight coloredeq eqj" style="">t</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span></span></span></span></span></span>、时间步<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.61508em;vertical-align:0em;"></span><span class="mord coloredeq eqj" style=""><span class="mord mathnormal" style="">t</span></span></span></span></span></span>长和条件环境预测噪声<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.43056em;vertical-align:0em;"></span><span class="mord coloredeq eqi" style=""><span class="mord mathnormal" style="">c</span></span></span></span></span></span>。</p>
|
||||
<p><span ><span class="katex-display"><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:1em;vertical-align:-0.25em;"></span><span class="mord coloredeq eqb" style=""><span class="mord" style=""><span class="mord mathnormal" style="">ϵ</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.33610799999999996em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord text mtight" style=""><span class="mord mtight" style=""><span class="mord mtight coloredeq eqi" style="">c</span></span><span class="mord mtight" style="">ond</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span><span class="mopen" style="">(</span><span class="mord" style=""><span class="mord coloredeq eqg" style=""><span class="mord mathnormal" style="">x</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.2805559999999999em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style=""><span class="mord mathnormal mtight coloredeq eqj" style="">t</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span></span><span class="mpunct" style="">,</span><span class="mspace" style="margin-right:0.16666666666666666em;"></span><span class="mord" style=""><span class="mord mathnormal coloredeq eqi" style="">c</span></span><span class="mclose" style="">)</span></span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">136</span> <span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span> <span class="n">t</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span> <span class="n">context</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-24'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-24'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">145</span> <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">model</span><span class="p">(</span><span class="n">x</span><span class="p">,</span> <span class="n">t</span><span class="p">,</span> <span class="n">context</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
File diff suppressed because it is too large
Load Diff
@@ -0,0 +1,248 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content="CLIP 嵌入器可获得提示性嵌入以实现稳定的扩散"/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="CLIP 文本嵌入器"/>
|
||||
<meta name="twitter:description" content="CLIP 嵌入器可获得提示性嵌入以实现稳定的扩散"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/diffusion/stable_diffusion/model/clip_embedder.html"/>
|
||||
<meta property="og:title" content="CLIP 文本嵌入器"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="CLIP 文本嵌入器"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="CLIP 文本嵌入器"/>
|
||||
<meta property="og:description" content="CLIP 嵌入器可获得提示性嵌入以实现稳定的扩散"/>
|
||||
|
||||
<title>CLIP 文本嵌入器</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../../../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/diffusion/stable_diffusion/model/clip_embedder.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="../../index.html">diffusion</a>
|
||||
<a class="parent" href="../index.html">stable_diffusion</a>
|
||||
<a class="parent" href="index.html">model</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/diffusion/stable_diffusion/model/clip_embedder.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1>CLIP 文本嵌入器</h1>
|
||||
<p>这用于获取提示嵌入以实现<a href="../index.html">稳定的扩散</a>。它使用 HuggingFace 变形金刚 CLIP 模型。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">14</span><span></span><span class="kn">from</span> <span class="nn">typing</span> <span class="kn">import</span> <span class="n">List</span>
|
||||
<span class="lineno">15</span>
|
||||
<span class="lineno">16</span><span class="kn">from</span> <span class="nn">torch</span> <span class="kn">import</span> <span class="n">nn</span>
|
||||
<span class="lineno">17</span><span class="kn">from</span> <span class="nn">transformers</span> <span class="kn">import</span> <span class="n">CLIPTokenizer</span><span class="p">,</span> <span class="n">CLIPTextModel</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-1'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-1'>#</a>
|
||||
</div>
|
||||
<h2>CLIP 文本嵌入器</h2>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">20</span><span class="k">class</span> <span class="nc">CLIPTextEmbedder</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-2'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-2'>#</a>
|
||||
</div>
|
||||
<ul><li><code class="highlight"><span></span><span class="n">version</span></code>
|
||||
是模型版本</li>
|
||||
<li><code class="highlight"><span></span><span class="n">device</span></code>
|
||||
是设备吗</li>
|
||||
<li><code class="highlight"><span></span><span class="n">max_length</span></code>
|
||||
是标记化提示的最大长度</li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">25</span> <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">version</span><span class="p">:</span> <span class="nb">str</span> <span class="o">=</span> <span class="s2">"openai/clip-vit-large-patch14"</span><span class="p">,</span> <span class="n">device</span><span class="o">=</span><span class="s2">"cuda:0"</span><span class="p">,</span> <span class="n">max_length</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">77</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-3'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-3'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">31</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-4'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-4'>#</a>
|
||||
</div>
|
||||
<p>加载代币生成器</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">33</span> <span class="bp">self</span><span class="o">.</span><span class="n">tokenizer</span> <span class="o">=</span> <span class="n">CLIPTokenizer</span><span class="o">.</span><span class="n">from_pretrained</span><span class="p">(</span><span class="n">version</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-5'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-5'>#</a>
|
||||
</div>
|
||||
<p>加载 CLIP 变压器</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">35</span> <span class="bp">self</span><span class="o">.</span><span class="n">transformer</span> <span class="o">=</span> <span class="n">CLIPTextModel</span><span class="o">.</span><span class="n">from_pretrained</span><span class="p">(</span><span class="n">version</span><span class="p">)</span><span class="o">.</span><span class="n">eval</span><span class="p">()</span>
|
||||
<span class="lineno">36</span>
|
||||
<span class="lineno">37</span> <span class="bp">self</span><span class="o">.</span><span class="n">device</span> <span class="o">=</span> <span class="n">device</span>
|
||||
<span class="lineno">38</span> <span class="bp">self</span><span class="o">.</span><span class="n">max_length</span> <span class="o">=</span> <span class="n">max_length</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-6'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-6'>#</a>
|
||||
</div>
|
||||
<ul><li><code class="highlight"><span></span><span class="n">prompts</span></code>
|
||||
是要嵌入的提示列表</li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">40</span> <span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">prompts</span><span class="p">:</span> <span class="n">List</span><span class="p">[</span><span class="nb">str</span><span class="p">]):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-7'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-7'>#</a>
|
||||
</div>
|
||||
<p>对提示进行标记化</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">45</span> <span class="n">batch_encoding</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">tokenizer</span><span class="p">(</span><span class="n">prompts</span><span class="p">,</span> <span class="n">truncation</span><span class="o">=</span><span class="kc">True</span><span class="p">,</span> <span class="n">max_length</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">max_length</span><span class="p">,</span> <span class="n">return_length</span><span class="o">=</span><span class="kc">True</span><span class="p">,</span>
|
||||
<span class="lineno">46</span> <span class="n">return_overflowing_tokens</span><span class="o">=</span><span class="kc">False</span><span class="p">,</span> <span class="n">padding</span><span class="o">=</span><span class="s2">"max_length"</span><span class="p">,</span> <span class="n">return_tensors</span><span class="o">=</span><span class="s2">"pt"</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-8'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-8'>#</a>
|
||||
</div>
|
||||
<p>获取代币 ID</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">48</span> <span class="n">tokens</span> <span class="o">=</span> <span class="n">batch_encoding</span><span class="p">[</span><span class="s2">"input_ids"</span><span class="p">]</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">device</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-9'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-9'>#</a>
|
||||
</div>
|
||||
<p>获取 CLIP 嵌入内容</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">50</span> <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">transformer</span><span class="p">(</span><span class="n">input_ids</span><span class="o">=</span><span class="n">tokens</span><span class="p">)</span><span class="o">.</span><span class="n">last_hidden_state</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../../../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
@@ -0,0 +1,130 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content="用于稳定扩散的模型和组件。"/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="用于稳定扩散的模块"/>
|
||||
<meta name="twitter:description" content="用于稳定扩散的模型和组件。"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/diffusion/stable_diffusion/model/index.html"/>
|
||||
<meta property="og:title" content="用于稳定扩散的模块"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="用于稳定扩散的模块"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="用于稳定扩散的模块"/>
|
||||
<meta property="og:description" content="用于稳定扩散的模型和组件。"/>
|
||||
|
||||
<title>用于稳定扩散的模块</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../../../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/diffusion/stable_diffusion/model/index.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="../../index.html">diffusion</a>
|
||||
<a class="parent" href="../index.html">stable_diffusion</a>
|
||||
<a class="parent" href="index.html">model</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/diffusion/stable_diffusion/model/__init__.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1><a href="../index.html">稳定的扩散</a>模型</h1>
|
||||
<ul><li><a href="autoencoder.html">自动编码器</a></li>
|
||||
<li><a href="unet_attention.html">备受关注</a>的 <a href="unet.html">U-Net</a></li>
|
||||
<li><a href="clip_embedder.html">CLIP 嵌入器</a>。</li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../../../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
File diff suppressed because one or more lines are too long
File diff suppressed because it is too large
Load Diff
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
@@ -0,0 +1,130 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content="带注释的 PyTorch 实现/稳定扩散示例用法教程"/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="用于显示示例用法和稳定扩散的脚本"/>
|
||||
<meta name="twitter:description" content="带注释的 PyTorch 实现/稳定扩散示例用法教程"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/diffusion/stable_diffusion/scripts/index.html"/>
|
||||
<meta property="og:title" content="用于显示示例用法和稳定扩散的脚本"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="用于显示示例用法和稳定扩散的脚本"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="用于显示示例用法和稳定扩散的脚本"/>
|
||||
<meta property="og:description" content="带注释的 PyTorch 实现/稳定扩散示例用法教程"/>
|
||||
|
||||
<title>用于显示示例用法和稳定扩散的脚本</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../../../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/diffusion/stable_diffusion/scripts/index.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="../../index.html">diffusion</a>
|
||||
<a class="parent" href="../index.html">stable_diffusion</a>
|
||||
<a class="parent" href="index.html">scripts</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/diffusion/stable_diffusion/scripts/__init__.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1>用于显示示例用法和<a href="../index.html">稳定扩散</a>的脚本</h1>
|
||||
<ul><li><a href="text_to_image.html">提示图像扩散</a></li>
|
||||
<li><a href="image_to_image.html">图像到图像的扩散</a></li>
|
||||
<li><a href="in_paint.html">内画</a></li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../../../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
File diff suppressed because one or more lines are too long
@@ -0,0 +1,478 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content="用于稳定扩散的实用函数"/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="用于稳定扩散的实用函数"/>
|
||||
<meta name="twitter:description" content="用于稳定扩散的实用函数"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/diffusion/stable_diffusion/util.html"/>
|
||||
<meta property="og:title" content="用于稳定扩散的实用函数"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="用于稳定扩散的实用函数"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="用于稳定扩散的实用函数"/>
|
||||
<meta property="og:description" content="用于稳定扩散的实用函数"/>
|
||||
|
||||
<title>用于稳定扩散的实用函数</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/diffusion/stable_diffusion/util.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="../index.html">diffusion</a>
|
||||
<a class="parent" href="index.html">stable_diffusion</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/diffusion/stable_diffusion/util.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1>用于<a href="index.html">稳定扩散</a>的实用函数</h1>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">11</span><span></span><span class="kn">import</span> <span class="nn">os</span>
|
||||
<span class="lineno">12</span><span class="kn">import</span> <span class="nn">random</span>
|
||||
<span class="lineno">13</span><span class="kn">from</span> <span class="nn">pathlib</span> <span class="kn">import</span> <span class="n">Path</span>
|
||||
<span class="lineno">14</span>
|
||||
<span class="lineno">15</span><span class="kn">import</span> <span class="nn">PIL</span>
|
||||
<span class="lineno">16</span><span class="kn">import</span> <span class="nn">numpy</span> <span class="k">as</span> <span class="nn">np</span>
|
||||
<span class="lineno">17</span><span class="kn">import</span> <span class="nn">torch</span>
|
||||
<span class="lineno">18</span><span class="kn">from</span> <span class="nn">PIL</span> <span class="kn">import</span> <span class="n">Image</span>
|
||||
<span class="lineno">19</span>
|
||||
<span class="lineno">20</span><span class="kn">from</span> <span class="nn">labml</span> <span class="kn">import</span> <span class="n">monit</span>
|
||||
<span class="lineno">21</span><span class="kn">from</span> <span class="nn">labml.logger</span> <span class="kn">import</span> <span class="n">inspect</span>
|
||||
<span class="lineno">22</span><span class="kn">from</span> <span class="nn">labml_nn.diffusion.stable_diffusion.latent_diffusion</span> <span class="kn">import</span> <span class="n">LatentDiffusion</span>
|
||||
<span class="lineno">23</span><span class="kn">from</span> <span class="nn">labml_nn.diffusion.stable_diffusion.model.autoencoder</span> <span class="kn">import</span> <span class="n">Encoder</span><span class="p">,</span> <span class="n">Decoder</span><span class="p">,</span> <span class="n">Autoencoder</span>
|
||||
<span class="lineno">24</span><span class="kn">from</span> <span class="nn">labml_nn.diffusion.stable_diffusion.model.clip_embedder</span> <span class="kn">import</span> <span class="n">CLIPTextEmbedder</span>
|
||||
<span class="lineno">25</span><span class="kn">from</span> <span class="nn">labml_nn.diffusion.stable_diffusion.model.unet</span> <span class="kn">import</span> <span class="n">UNetModel</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-1'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-1'>#</a>
|
||||
</div>
|
||||
<h3>设置随机种子</h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">28</span><span class="k">def</span> <span class="nf">set_seed</span><span class="p">(</span><span class="n">seed</span><span class="p">:</span> <span class="nb">int</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-2'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-2'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">32</span> <span class="n">random</span><span class="o">.</span><span class="n">seed</span><span class="p">(</span><span class="n">seed</span><span class="p">)</span>
|
||||
<span class="lineno">33</span> <span class="n">np</span><span class="o">.</span><span class="n">random</span><span class="o">.</span><span class="n">seed</span><span class="p">(</span><span class="n">seed</span><span class="p">)</span>
|
||||
<span class="lineno">34</span> <span class="n">torch</span><span class="o">.</span><span class="n">manual_seed</span><span class="p">(</span><span class="n">seed</span><span class="p">)</span>
|
||||
<span class="lineno">35</span> <span class="n">torch</span><span class="o">.</span><span class="n">cuda</span><span class="o">.</span><span class="n">manual_seed_all</span><span class="p">(</span><span class="n">seed</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-3'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-3'>#</a>
|
||||
</div>
|
||||
<h3>加载<a href="latent_diffusion.html"><code class="highlight"><span></span><span class="n">LatentDiffusion</span></code>
|
||||
模型</a></h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">38</span><span class="k">def</span> <span class="nf">load_model</span><span class="p">(</span><span class="n">path</span><span class="p">:</span> <span class="n">Path</span> <span class="o">=</span> <span class="kc">None</span><span class="p">)</span> <span class="o">-></span> <span class="n">LatentDiffusion</span><span class="p">:</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-4'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-4'>#</a>
|
||||
</div>
|
||||
<p>初始化自动编码器</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">44</span> <span class="k">with</span> <span class="n">monit</span><span class="o">.</span><span class="n">section</span><span class="p">(</span><span class="s1">'Initialize autoencoder'</span><span class="p">):</span>
|
||||
<span class="lineno">45</span> <span class="n">encoder</span> <span class="o">=</span> <span class="n">Encoder</span><span class="p">(</span><span class="n">z_channels</span><span class="o">=</span><span class="mi">4</span><span class="p">,</span>
|
||||
<span class="lineno">46</span> <span class="n">in_channels</span><span class="o">=</span><span class="mi">3</span><span class="p">,</span>
|
||||
<span class="lineno">47</span> <span class="n">channels</span><span class="o">=</span><span class="mi">128</span><span class="p">,</span>
|
||||
<span class="lineno">48</span> <span class="n">channel_multipliers</span><span class="o">=</span><span class="p">[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="mi">4</span><span class="p">,</span> <span class="mi">4</span><span class="p">],</span>
|
||||
<span class="lineno">49</span> <span class="n">n_resnet_blocks</span><span class="o">=</span><span class="mi">2</span><span class="p">)</span>
|
||||
<span class="lineno">50</span>
|
||||
<span class="lineno">51</span> <span class="n">decoder</span> <span class="o">=</span> <span class="n">Decoder</span><span class="p">(</span><span class="n">out_channels</span><span class="o">=</span><span class="mi">3</span><span class="p">,</span>
|
||||
<span class="lineno">52</span> <span class="n">z_channels</span><span class="o">=</span><span class="mi">4</span><span class="p">,</span>
|
||||
<span class="lineno">53</span> <span class="n">channels</span><span class="o">=</span><span class="mi">128</span><span class="p">,</span>
|
||||
<span class="lineno">54</span> <span class="n">channel_multipliers</span><span class="o">=</span><span class="p">[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="mi">4</span><span class="p">,</span> <span class="mi">4</span><span class="p">],</span>
|
||||
<span class="lineno">55</span> <span class="n">n_resnet_blocks</span><span class="o">=</span><span class="mi">2</span><span class="p">)</span>
|
||||
<span class="lineno">56</span>
|
||||
<span class="lineno">57</span> <span class="n">autoencoder</span> <span class="o">=</span> <span class="n">Autoencoder</span><span class="p">(</span><span class="n">emb_channels</span><span class="o">=</span><span class="mi">4</span><span class="p">,</span>
|
||||
<span class="lineno">58</span> <span class="n">encoder</span><span class="o">=</span><span class="n">encoder</span><span class="p">,</span>
|
||||
<span class="lineno">59</span> <span class="n">decoder</span><span class="o">=</span><span class="n">decoder</span><span class="p">,</span>
|
||||
<span class="lineno">60</span> <span class="n">z_channels</span><span class="o">=</span><span class="mi">4</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-5'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-5'>#</a>
|
||||
</div>
|
||||
<p>初始化 CLIP 文本嵌入器</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">63</span> <span class="k">with</span> <span class="n">monit</span><span class="o">.</span><span class="n">section</span><span class="p">(</span><span class="s1">'Initialize CLIP Embedder'</span><span class="p">):</span>
|
||||
<span class="lineno">64</span> <span class="n">clip_text_embedder</span> <span class="o">=</span> <span class="n">CLIPTextEmbedder</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-6'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-6'>#</a>
|
||||
</div>
|
||||
<p>初始化 U-Net</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">67</span> <span class="k">with</span> <span class="n">monit</span><span class="o">.</span><span class="n">section</span><span class="p">(</span><span class="s1">'Initialize U-Net'</span><span class="p">):</span>
|
||||
<span class="lineno">68</span> <span class="n">unet_model</span> <span class="o">=</span> <span class="n">UNetModel</span><span class="p">(</span><span class="n">in_channels</span><span class="o">=</span><span class="mi">4</span><span class="p">,</span>
|
||||
<span class="lineno">69</span> <span class="n">out_channels</span><span class="o">=</span><span class="mi">4</span><span class="p">,</span>
|
||||
<span class="lineno">70</span> <span class="n">channels</span><span class="o">=</span><span class="mi">320</span><span class="p">,</span>
|
||||
<span class="lineno">71</span> <span class="n">attention_levels</span><span class="o">=</span><span class="p">[</span><span class="mi">0</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">2</span><span class="p">],</span>
|
||||
<span class="lineno">72</span> <span class="n">n_res_blocks</span><span class="o">=</span><span class="mi">2</span><span class="p">,</span>
|
||||
<span class="lineno">73</span> <span class="n">channel_multipliers</span><span class="o">=</span><span class="p">[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="mi">4</span><span class="p">,</span> <span class="mi">4</span><span class="p">],</span>
|
||||
<span class="lineno">74</span> <span class="n">n_heads</span><span class="o">=</span><span class="mi">8</span><span class="p">,</span>
|
||||
<span class="lineno">75</span> <span class="n">tf_layers</span><span class="o">=</span><span class="mi">1</span><span class="p">,</span>
|
||||
<span class="lineno">76</span> <span class="n">d_cond</span><span class="o">=</span><span class="mi">768</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-7'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-7'>#</a>
|
||||
</div>
|
||||
<p>初始化潜在扩散模型</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">79</span> <span class="k">with</span> <span class="n">monit</span><span class="o">.</span><span class="n">section</span><span class="p">(</span><span class="s1">'Initialize Latent Diffusion model'</span><span class="p">):</span>
|
||||
<span class="lineno">80</span> <span class="n">model</span> <span class="o">=</span> <span class="n">LatentDiffusion</span><span class="p">(</span><span class="n">linear_start</span><span class="o">=</span><span class="mf">0.00085</span><span class="p">,</span>
|
||||
<span class="lineno">81</span> <span class="n">linear_end</span><span class="o">=</span><span class="mf">0.0120</span><span class="p">,</span>
|
||||
<span class="lineno">82</span> <span class="n">n_steps</span><span class="o">=</span><span class="mi">1000</span><span class="p">,</span>
|
||||
<span class="lineno">83</span> <span class="n">latent_scaling_factor</span><span class="o">=</span><span class="mf">0.18215</span><span class="p">,</span>
|
||||
<span class="lineno">84</span>
|
||||
<span class="lineno">85</span> <span class="n">autoencoder</span><span class="o">=</span><span class="n">autoencoder</span><span class="p">,</span>
|
||||
<span class="lineno">86</span> <span class="n">clip_embedder</span><span class="o">=</span><span class="n">clip_text_embedder</span><span class="p">,</span>
|
||||
<span class="lineno">87</span> <span class="n">unet_model</span><span class="o">=</span><span class="n">unet_model</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-8'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-8'>#</a>
|
||||
</div>
|
||||
<p>加载检查点</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">90</span> <span class="k">with</span> <span class="n">monit</span><span class="o">.</span><span class="n">section</span><span class="p">(</span><span class="sa">f</span><span class="s2">"Loading model from </span><span class="si">{</span><span class="n">path</span><span class="si">}</span><span class="s2">"</span><span class="p">):</span>
|
||||
<span class="lineno">91</span> <span class="n">checkpoint</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">load</span><span class="p">(</span><span class="n">path</span><span class="p">,</span> <span class="n">map_location</span><span class="o">=</span><span class="s2">"cpu"</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-9'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-9'>#</a>
|
||||
</div>
|
||||
<p>设置模型状态</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">94</span> <span class="k">with</span> <span class="n">monit</span><span class="o">.</span><span class="n">section</span><span class="p">(</span><span class="s1">'Load state'</span><span class="p">):</span>
|
||||
<span class="lineno">95</span> <span class="n">missing_keys</span><span class="p">,</span> <span class="n">extra_keys</span> <span class="o">=</span> <span class="n">model</span><span class="o">.</span><span class="n">load_state_dict</span><span class="p">(</span><span class="n">checkpoint</span><span class="p">[</span><span class="s2">"state_dict"</span><span class="p">],</span> <span class="n">strict</span><span class="o">=</span><span class="kc">False</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-10'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-10'>#</a>
|
||||
</div>
|
||||
<p>调试输出</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">98</span> <span class="n">inspect</span><span class="p">(</span><span class="n">global_step</span><span class="o">=</span><span class="n">checkpoint</span><span class="o">.</span><span class="n">get</span><span class="p">(</span><span class="s1">'global_step'</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">),</span> <span class="n">missing_keys</span><span class="o">=</span><span class="n">missing_keys</span><span class="p">,</span> <span class="n">extra_keys</span><span class="o">=</span><span class="n">extra_keys</span><span class="p">,</span>
|
||||
<span class="lineno">99</span> <span class="n">_expand</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-11'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-11'>#</a>
|
||||
</div>
|
||||
<p></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">102</span> <span class="n">model</span><span class="o">.</span><span class="n">eval</span><span class="p">()</span>
|
||||
<span class="lineno">103</span> <span class="k">return</span> <span class="n">model</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-12'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-12'>#</a>
|
||||
</div>
|
||||
<h3>加载图片</h3>
|
||||
<p>这将从文件加载图像并返回 PyTorch 张量。</p>
|
||||
<ul><li><code class="highlight"><span></span><span class="n">path</span></code>
|
||||
是图像的路径</li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">106</span><span class="k">def</span> <span class="nf">load_img</span><span class="p">(</span><span class="n">path</span><span class="p">:</span> <span class="nb">str</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-13'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-13'>#</a>
|
||||
</div>
|
||||
<p>打开图片</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">115</span> <span class="n">image</span> <span class="o">=</span> <span class="n">Image</span><span class="o">.</span><span class="n">open</span><span class="p">(</span><span class="n">path</span><span class="p">)</span><span class="o">.</span><span class="n">convert</span><span class="p">(</span><span class="s2">"RGB"</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-14'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-14'>#</a>
|
||||
</div>
|
||||
<p>获取图像大小</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">117</span> <span class="n">w</span><span class="p">,</span> <span class="n">h</span> <span class="o">=</span> <span class="n">image</span><span class="o">.</span><span class="n">size</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-15'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-15'>#</a>
|
||||
</div>
|
||||
<p>调整为 32 的倍数</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">119</span> <span class="n">w</span> <span class="o">=</span> <span class="n">w</span> <span class="o">-</span> <span class="n">w</span> <span class="o">%</span> <span class="mi">32</span>
|
||||
<span class="lineno">120</span> <span class="n">h</span> <span class="o">=</span> <span class="n">h</span> <span class="o">-</span> <span class="n">h</span> <span class="o">%</span> <span class="mi">32</span>
|
||||
<span class="lineno">121</span> <span class="n">image</span> <span class="o">=</span> <span class="n">image</span><span class="o">.</span><span class="n">resize</span><span class="p">((</span><span class="n">w</span><span class="p">,</span> <span class="n">h</span><span class="p">),</span> <span class="n">resample</span><span class="o">=</span><span class="n">PIL</span><span class="o">.</span><span class="n">Image</span><span class="o">.</span><span class="n">LANCZOS</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-16'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-16'>#</a>
|
||||
</div>
|
||||
<p>转换为 numpy 并映射到 fo<code class="highlight"><span></span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">]</span></code>
|
||||
r<code class="highlight"><span></span><span class="p">[</span><span class="mi">0</span><span class="p">,</span> <span class="mi">255</span><span class="p">]</span></code>
|
||||
</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">123</span> <span class="n">image</span> <span class="o">=</span> <span class="n">np</span><span class="o">.</span><span class="n">array</span><span class="p">(</span><span class="n">image</span><span class="p">)</span><span class="o">.</span><span class="n">astype</span><span class="p">(</span><span class="n">np</span><span class="o">.</span><span class="n">float32</span><span class="p">)</span> <span class="o">*</span> <span class="p">(</span><span class="mf">2.</span> <span class="o">/</span> <span class="mf">255.0</span><span class="p">)</span> <span class="o">-</span> <span class="mi">1</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-17'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-17'>#</a>
|
||||
</div>
|
||||
<p>转置成形状<code class="highlight"><span></span><span class="p">[</span><span class="n">batch_size</span><span class="p">,</span> <span class="n">channels</span><span class="p">,</span> <span class="n">height</span><span class="p">,</span> <span class="n">width</span><span class="p">]</span></code>
|
||||
</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">125</span> <span class="n">image</span> <span class="o">=</span> <span class="n">image</span><span class="p">[</span><span class="kc">None</span><span class="p">]</span><span class="o">.</span><span class="n">transpose</span><span class="p">(</span><span class="mi">0</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">2</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-18'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-18'>#</a>
|
||||
</div>
|
||||
<p>转换为 torch</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">127</span> <span class="k">return</span> <span class="n">torch</span><span class="o">.</span><span class="n">from_numpy</span><span class="p">(</span><span class="n">image</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-19'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-19'>#</a>
|
||||
</div>
|
||||
<h3>保存图像</h3>
|
||||
<ul><li><code class="highlight"><span></span><span class="n">images</span></code>
|
||||
是带有形状图像的张量<code class="highlight"><span></span><span class="p">[</span><span class="n">batch_size</span><span class="p">,</span> <span class="n">channels</span><span class="p">,</span> <span class="n">height</span><span class="p">,</span> <span class="n">width</span><span class="p">]</span></code>
|
||||
</li>
|
||||
<li><code class="highlight"><span></span><span class="n">dest_path</span></code>
|
||||
是保存图像的文件夹</li>
|
||||
<li><code class="highlight"><span></span><span class="n">prefix</span></code>
|
||||
是添加到文件名的前缀</li>
|
||||
<li><code class="highlight"><span></span><span class="n">img_format</span></code>
|
||||
是图像格式</li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">130</span><span class="k">def</span> <span class="nf">save_images</span><span class="p">(</span><span class="n">images</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span> <span class="n">dest_path</span><span class="p">:</span> <span class="nb">str</span><span class="p">,</span> <span class="n">prefix</span><span class="p">:</span> <span class="nb">str</span> <span class="o">=</span> <span class="s1">''</span><span class="p">,</span> <span class="n">img_format</span><span class="p">:</span> <span class="nb">str</span> <span class="o">=</span> <span class="s1">'jpeg'</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-20'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-20'>#</a>
|
||||
</div>
|
||||
<p>创建目标文件夹</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">141</span> <span class="n">os</span><span class="o">.</span><span class="n">makedirs</span><span class="p">(</span><span class="n">dest_path</span><span class="p">,</span> <span class="n">exist_ok</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-21'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-21'>#</a>
|
||||
</div>
|
||||
<p>将图像映射到<code class="highlight"><span></span><span class="p">[</span><span class="mi">0</span><span class="p">,</span> <span class="mi">1</span><span class="p">]</span></code>
|
||||
空间并剪辑</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">144</span> <span class="n">images</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">clamp</span><span class="p">((</span><span class="n">images</span> <span class="o">+</span> <span class="mf">1.0</span><span class="p">)</span> <span class="o">/</span> <span class="mf">2.0</span><span class="p">,</span> <span class="nb">min</span><span class="o">=</span><span class="mf">0.0</span><span class="p">,</span> <span class="nb">max</span><span class="o">=</span><span class="mf">1.0</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-22'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-22'>#</a>
|
||||
</div>
|
||||
<p>转置为 numpy<code class="highlight"><span></span><span class="p">[</span><span class="n">batch_size</span><span class="p">,</span> <span class="n">height</span><span class="p">,</span> <span class="n">width</span><span class="p">,</span> <span class="n">channels</span><span class="p">]</span></code>
|
||||
并转换为 numpy</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">146</span> <span class="n">images</span> <span class="o">=</span> <span class="n">images</span><span class="o">.</span><span class="n">cpu</span><span class="p">()</span><span class="o">.</span><span class="n">permute</span><span class="p">(</span><span class="mi">0</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span><span class="o">.</span><span class="n">numpy</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-23'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-23'>#</a>
|
||||
</div>
|
||||
<p>保存图片</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">149</span> <span class="k">for</span> <span class="n">i</span><span class="p">,</span> <span class="n">img</span> <span class="ow">in</span> <span class="nb">enumerate</span><span class="p">(</span><span class="n">images</span><span class="p">):</span>
|
||||
<span class="lineno">150</span> <span class="n">img</span> <span class="o">=</span> <span class="n">Image</span><span class="o">.</span><span class="n">fromarray</span><span class="p">((</span><span class="mf">255.</span> <span class="o">*</span> <span class="n">img</span><span class="p">)</span><span class="o">.</span><span class="n">astype</span><span class="p">(</span><span class="n">np</span><span class="o">.</span><span class="n">uint8</span><span class="p">))</span>
|
||||
<span class="lineno">151</span> <span class="n">img</span><span class="o">.</span><span class="n">save</span><span class="p">(</span><span class="n">os</span><span class="o">.</span><span class="n">path</span><span class="o">.</span><span class="n">join</span><span class="p">(</span><span class="n">dest_path</span><span class="p">,</span> <span class="sa">f</span><span class="s2">"</span><span class="si">{</span><span class="n">prefix</span><span class="si">}{</span><span class="n">i</span><span class="si">:</span><span class="s2">05</span><span class="si">}</span><span class="s2">.</span><span class="si">{</span><span class="n">img_format</span><span class="si">}</span><span class="s2">"</span><span class="p">),</span> <span class="nb">format</span><span class="o">=</span><span class="n">img_format</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
File diff suppressed because one or more lines are too long
@@ -0,0 +1,390 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content="在 CIFAR 10 上训练一个大型模型进行蒸馏。"/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="在 CIFAR 10 上训练一个大型模型"/>
|
||||
<meta name="twitter:description" content="在 CIFAR 10 上训练一个大型模型进行蒸馏。"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/distillation/large.html"/>
|
||||
<meta property="og:title" content="在 CIFAR 10 上训练一个大型模型"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="在 CIFAR 10 上训练一个大型模型"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="在 CIFAR 10 上训练一个大型模型"/>
|
||||
<meta property="og:description" content="在 CIFAR 10 上训练一个大型模型进行蒸馏。"/>
|
||||
|
||||
<title>在 CIFAR 10 上训练一个大型模型</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/distillation/large.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="index.html">distillation</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/distillation/large.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1>在 CIFAR 10 上训练一个大型模型</h1>
|
||||
<p>这会在 CIFAR 10 上训练一个用于<a href="index.html">蒸馏</a>的大型模型。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">13</span><span></span><span class="kn">import</span> <span class="nn">torch.nn</span> <span class="k">as</span> <span class="nn">nn</span>
|
||||
<span class="lineno">14</span>
|
||||
<span class="lineno">15</span><span class="kn">from</span> <span class="nn">labml</span> <span class="kn">import</span> <span class="n">experiment</span><span class="p">,</span> <span class="n">logger</span>
|
||||
<span class="lineno">16</span><span class="kn">from</span> <span class="nn">labml.configs</span> <span class="kn">import</span> <span class="n">option</span>
|
||||
<span class="lineno">17</span><span class="kn">from</span> <span class="nn">labml_nn.experiments.cifar10</span> <span class="kn">import</span> <span class="n">CIFAR10Configs</span><span class="p">,</span> <span class="n">CIFAR10VGGModel</span>
|
||||
<span class="lineno">18</span><span class="kn">from</span> <span class="nn">labml_nn.normalization.batch_norm</span> <span class="kn">import</span> <span class="n">BatchNorm</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-1'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-1'>#</a>
|
||||
</div>
|
||||
<h2>配置</h2>
|
||||
<p>我们使用<a href="../experiments/cifar10.html"><code class="highlight"><span></span><span class="n">CIFAR10Configs</span></code>
|
||||
</a>它来定义所有与数据集相关的配置、优化器和训练循环。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">21</span><span class="k">class</span> <span class="nc">Configs</span><span class="p">(</span><span class="n">CIFAR10Configs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-2'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-2'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">28</span> <span class="k">pass</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-3'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-3'>#</a>
|
||||
</div>
|
||||
<h3>适用于 CIFAR-10 分类的 VGG 样式模型</h3>
|
||||
<p>这源于<a href="../experiments/cifar10.html">通用的 VGG 风格架构</a>。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">31</span><span class="k">class</span> <span class="nc">LargeModel</span><span class="p">(</span><span class="n">CIFAR10VGGModel</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-4'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-4'>#</a>
|
||||
</div>
|
||||
<p>创建卷积层和激活</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">38</span> <span class="k">def</span> <span class="nf">conv_block</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">in_channels</span><span class="p">,</span> <span class="n">out_channels</span><span class="p">)</span> <span class="o">-></span> <span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">:</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-5'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-5'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">42</span> <span class="k">return</span> <span class="n">nn</span><span class="o">.</span><span class="n">Sequential</span><span class="p">(</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-6'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-6'>#</a>
|
||||
</div>
|
||||
<p>辍学</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">44</span> <span class="n">nn</span><span class="o">.</span><span class="n">Dropout</span><span class="p">(</span><span class="mf">0.1</span><span class="p">),</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-7'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-7'>#</a>
|
||||
</div>
|
||||
<p>卷积层</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">46</span> <span class="n">nn</span><span class="o">.</span><span class="n">Conv2d</span><span class="p">(</span><span class="n">in_channels</span><span class="p">,</span> <span class="n">out_channels</span><span class="p">,</span> <span class="n">kernel_size</span><span class="o">=</span><span class="mi">3</span><span class="p">,</span> <span class="n">padding</span><span class="o">=</span><span class="mi">1</span><span class="p">),</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-8'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-8'>#</a>
|
||||
</div>
|
||||
<p>批量标准化</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">48</span> <span class="n">BatchNorm</span><span class="p">(</span><span class="n">out_channels</span><span class="p">,</span> <span class="n">track_running_stats</span><span class="o">=</span><span class="kc">False</span><span class="p">),</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-9'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-9'>#</a>
|
||||
</div>
|
||||
<p>激活 ReLU</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">50</span> <span class="n">nn</span><span class="o">.</span><span class="n">ReLU</span><span class="p">(</span><span class="n">inplace</span><span class="o">=</span><span class="kc">True</span><span class="p">),</span>
|
||||
<span class="lineno">51</span> <span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-10'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-10'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">53</span> <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-11'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-11'>#</a>
|
||||
</div>
|
||||
<p>使用给定的卷积大小（通道）创建模型</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">55</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">([[</span><span class="mi">64</span><span class="p">,</span> <span class="mi">64</span><span class="p">],</span> <span class="p">[</span><span class="mi">128</span><span class="p">,</span> <span class="mi">128</span><span class="p">],</span> <span class="p">[</span><span class="mi">256</span><span class="p">,</span> <span class="mi">256</span><span class="p">,</span> <span class="mi">256</span><span class="p">],</span> <span class="p">[</span><span class="mi">512</span><span class="p">,</span> <span class="mi">512</span><span class="p">,</span> <span class="mi">512</span><span class="p">],</span> <span class="p">[</span><span class="mi">512</span><span class="p">,</span> <span class="mi">512</span><span class="p">,</span> <span class="mi">512</span><span class="p">]])</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-12'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-12'>#</a>
|
||||
</div>
|
||||
<h3>创建模型</h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">58</span><span class="nd">@option</span><span class="p">(</span><span class="n">Configs</span><span class="o">.</span><span class="n">model</span><span class="p">)</span>
|
||||
<span class="lineno">59</span><span class="k">def</span> <span class="nf">_large_model</span><span class="p">(</span><span class="n">c</span><span class="p">:</span> <span class="n">Configs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-13'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-13'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">63</span> <span class="k">return</span> <span class="n">LargeModel</span><span class="p">()</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">device</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-14'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-14'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">66</span><span class="k">def</span> <span class="nf">main</span><span class="p">():</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-15'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-15'>#</a>
|
||||
</div>
|
||||
<p>创建实验</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">68</span> <span class="n">experiment</span><span class="o">.</span><span class="n">create</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">'cifar10'</span><span class="p">,</span> <span class="n">comment</span><span class="o">=</span><span class="s1">'large model'</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-16'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-16'>#</a>
|
||||
</div>
|
||||
<p>创建配置</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">70</span> <span class="n">conf</span> <span class="o">=</span> <span class="n">Configs</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-17'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-17'>#</a>
|
||||
</div>
|
||||
<p>装载配置</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">72</span> <span class="n">experiment</span><span class="o">.</span><span class="n">configs</span><span class="p">(</span><span class="n">conf</span><span class="p">,</span> <span class="p">{</span>
|
||||
<span class="lineno">73</span> <span class="s1">'optimizer.optimizer'</span><span class="p">:</span> <span class="s1">'Adam'</span><span class="p">,</span>
|
||||
<span class="lineno">74</span> <span class="s1">'optimizer.learning_rate'</span><span class="p">:</span> <span class="mf">2.5e-4</span><span class="p">,</span>
|
||||
<span class="lineno">75</span> <span class="s1">'is_save_models'</span><span class="p">:</span> <span class="kc">True</span><span class="p">,</span>
|
||||
<span class="lineno">76</span> <span class="s1">'epochs'</span><span class="p">:</span> <span class="mi">20</span><span class="p">,</span>
|
||||
<span class="lineno">77</span> <span class="p">})</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-18'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-18'>#</a>
|
||||
</div>
|
||||
<p>设置保存/加载的模型</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">79</span> <span class="n">experiment</span><span class="o">.</span><span class="n">add_pytorch_models</span><span class="p">({</span><span class="s1">'model'</span><span class="p">:</span> <span class="n">conf</span><span class="o">.</span><span class="n">model</span><span class="p">})</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-19'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-19'>#</a>
|
||||
</div>
|
||||
<p>打印模型中参数的数量</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">81</span> <span class="n">logger</span><span class="o">.</span><span class="n">inspect</span><span class="p">(</span><span class="n">params</span><span class="o">=</span><span class="p">(</span><span class="nb">sum</span><span class="p">(</span><span class="n">p</span><span class="o">.</span><span class="n">numel</span><span class="p">()</span> <span class="k">for</span> <span class="n">p</span> <span class="ow">in</span> <span class="n">conf</span><span class="o">.</span><span class="n">model</span><span class="o">.</span><span class="n">parameters</span><span class="p">()</span> <span class="k">if</span> <span class="n">p</span><span class="o">.</span><span class="n">requires_grad</span><span class="p">)))</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-20'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-20'>#</a>
|
||||
</div>
|
||||
<p>开始实验并运行训练循环</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">83</span> <span class="k">with</span> <span class="n">experiment</span><span class="o">.</span><span class="n">start</span><span class="p">():</span>
|
||||
<span class="lineno">84</span> <span class="n">conf</span><span class="o">.</span><span class="n">run</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-21'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-21'>#</a>
|
||||
</div>
|
||||
<p></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">88</span><span class="k">if</span> <span class="vm">__name__</span> <span class="o">==</span> <span class="s1">'__main__'</span><span class="p">:</span>
|
||||
<span class="lineno">89</span> <span class="n">main</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
@@ -0,0 +1,129 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content=""/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="在神经网络中提炼知识"/>
|
||||
<meta name="twitter:description" content=""/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/distillation/readme.html"/>
|
||||
<meta property="og:title" content="在神经网络中提炼知识"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="在神经网络中提炼知识"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="在神经网络中提炼知识"/>
|
||||
<meta property="og:description" content=""/>
|
||||
|
||||
<title>在神经网络中提炼知识</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/distillation/readme.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="index.html">distillation</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/distillation/readme.md" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1><a href="https://nn.labml.ai/distillation/index.html">在神经网络中提炼知识</a></h1>
|
||||
<p>这是论文《<a href="https://arxiv.org/abs/1503.02531">在神经网络中提炼知识》的 PyT</a> <a href="https://pytorch.org">orch</a> 实现/教程。</p>
|
||||
<p>这是一种使用经过训练的大型网络中的知识来训练小型网络的方法；即从大型网络中提炼知识。</p>
|
||||
<p>直接在数据和标签上训练时，具有正则化或模型集合（使用 dropout）的大型模型比小型模型的概化效果更好。但是，在大型模型的帮助下，可以训练小模型以更好地进行概括。较小的模型在生产中更好：速度更快、计算更少、内存更少。</p>
|
||||
<p>经过训练的模型的输出概率比标签提供的信息更多，因为它也会为错误的类分配非零概率。这些概率告诉我们，样本有可能属于某些类别。例如，在对数字进行分类时，当给定数字 <em>7</em> 的图像时，广义模型会给出7的高概率，给2的概率很小但不是零，而给其他数字分配几乎为零的概率。蒸馏利用这些信息来更好地训练小型模型。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
@@ -0,0 +1,376 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content="在 CIFAR 10 上训练一个小模型，以测试蒸馏的益处。"/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="在 CIFAR 10 上训练一个小型模型"/>
|
||||
<meta name="twitter:description" content="在 CIFAR 10 上训练一个小模型，以测试蒸馏的益处。"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/distillation/small.html"/>
|
||||
<meta property="og:title" content="在 CIFAR 10 上训练一个小型模型"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="在 CIFAR 10 上训练一个小型模型"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="在 CIFAR 10 上训练一个小型模型"/>
|
||||
<meta property="og:description" content="在 CIFAR 10 上训练一个小模型，以测试蒸馏的益处。"/>
|
||||
|
||||
<title>在 CIFAR 10 上训练一个小型模型</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/distillation/small.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="index.html">distillation</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/distillation/small.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1>在 CIFAR 10 上训练一个小型模型</h1>
|
||||
<p>这在 CIFAR 10 上训练了一个小型模型，以测试<a href="index.html">蒸馏</a>的益处有多大。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">13</span><span></span><span class="kn">import</span> <span class="nn">torch.nn</span> <span class="k">as</span> <span class="nn">nn</span>
|
||||
<span class="lineno">14</span>
|
||||
<span class="lineno">15</span><span class="kn">from</span> <span class="nn">labml</span> <span class="kn">import</span> <span class="n">experiment</span><span class="p">,</span> <span class="n">logger</span>
|
||||
<span class="lineno">16</span><span class="kn">from</span> <span class="nn">labml.configs</span> <span class="kn">import</span> <span class="n">option</span>
|
||||
<span class="lineno">17</span><span class="kn">from</span> <span class="nn">labml_nn.experiments.cifar10</span> <span class="kn">import</span> <span class="n">CIFAR10Configs</span><span class="p">,</span> <span class="n">CIFAR10VGGModel</span>
|
||||
<span class="lineno">18</span><span class="kn">from</span> <span class="nn">labml_nn.normalization.batch_norm</span> <span class="kn">import</span> <span class="n">BatchNorm</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-1'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-1'>#</a>
|
||||
</div>
|
||||
<h2>配置</h2>
|
||||
<p>我们使用<a href="../experiments/cifar10.html"><code class="highlight"><span></span><span class="n">CIFAR10Configs</span></code>
|
||||
</a>它来定义所有与数据集相关的配置、优化器和训练循环。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">21</span><span class="k">class</span> <span class="nc">Configs</span><span class="p">(</span><span class="n">CIFAR10Configs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-2'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-2'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">28</span> <span class="k">pass</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-3'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-3'>#</a>
|
||||
</div>
|
||||
<h3>适用于 CIFAR-10 分类的 VGG 样式模型</h3>
|
||||
<p>这源于<a href="../experiments/cifar10.html">通用的 VGG 风格架构</a>。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">31</span><span class="k">class</span> <span class="nc">SmallModel</span><span class="p">(</span><span class="n">CIFAR10VGGModel</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-4'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-4'>#</a>
|
||||
</div>
|
||||
<p>创建卷积层和激活</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">38</span> <span class="k">def</span> <span class="nf">conv_block</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">in_channels</span><span class="p">,</span> <span class="n">out_channels</span><span class="p">)</span> <span class="o">-></span> <span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">:</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-5'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-5'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">42</span> <span class="k">return</span> <span class="n">nn</span><span class="o">.</span><span class="n">Sequential</span><span class="p">(</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-6'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-6'>#</a>
|
||||
</div>
|
||||
<p>卷积层</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">44</span> <span class="n">nn</span><span class="o">.</span><span class="n">Conv2d</span><span class="p">(</span><span class="n">in_channels</span><span class="p">,</span> <span class="n">out_channels</span><span class="p">,</span> <span class="n">kernel_size</span><span class="o">=</span><span class="mi">3</span><span class="p">,</span> <span class="n">padding</span><span class="o">=</span><span class="mi">1</span><span class="p">),</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-7'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-7'>#</a>
|
||||
</div>
|
||||
<p>批量标准化</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">46</span> <span class="n">BatchNorm</span><span class="p">(</span><span class="n">out_channels</span><span class="p">,</span> <span class="n">track_running_stats</span><span class="o">=</span><span class="kc">False</span><span class="p">),</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-8'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-8'>#</a>
|
||||
</div>
|
||||
<p>激活 ReLU</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">48</span> <span class="n">nn</span><span class="o">.</span><span class="n">ReLU</span><span class="p">(</span><span class="n">inplace</span><span class="o">=</span><span class="kc">True</span><span class="p">),</span>
|
||||
<span class="lineno">49</span> <span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-9'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-9'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">51</span> <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-10'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-10'>#</a>
|
||||
</div>
|
||||
<p>使用给定的卷积大小（通道）创建模型</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">53</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">([[</span><span class="mi">32</span><span class="p">,</span> <span class="mi">32</span><span class="p">],</span> <span class="p">[</span><span class="mi">64</span><span class="p">,</span> <span class="mi">64</span><span class="p">],</span> <span class="p">[</span><span class="mi">128</span><span class="p">],</span> <span class="p">[</span><span class="mi">128</span><span class="p">],</span> <span class="p">[</span><span class="mi">128</span><span class="p">]])</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-11'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-11'>#</a>
|
||||
</div>
|
||||
<h3>创建模型</h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">56</span><span class="nd">@option</span><span class="p">(</span><span class="n">Configs</span><span class="o">.</span><span class="n">model</span><span class="p">)</span>
|
||||
<span class="lineno">57</span><span class="k">def</span> <span class="nf">_small_model</span><span class="p">(</span><span class="n">c</span><span class="p">:</span> <span class="n">Configs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-12'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-12'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">61</span> <span class="k">return</span> <span class="n">SmallModel</span><span class="p">()</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">device</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-13'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-13'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">64</span><span class="k">def</span> <span class="nf">main</span><span class="p">():</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-14'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-14'>#</a>
|
||||
</div>
|
||||
<p>创建实验</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">66</span> <span class="n">experiment</span><span class="o">.</span><span class="n">create</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">'cifar10'</span><span class="p">,</span> <span class="n">comment</span><span class="o">=</span><span class="s1">'small model'</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-15'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-15'>#</a>
|
||||
</div>
|
||||
<p>创建配置</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">68</span> <span class="n">conf</span> <span class="o">=</span> <span class="n">Configs</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-16'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-16'>#</a>
|
||||
</div>
|
||||
<p>装载配置</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">70</span> <span class="n">experiment</span><span class="o">.</span><span class="n">configs</span><span class="p">(</span><span class="n">conf</span><span class="p">,</span> <span class="p">{</span>
|
||||
<span class="lineno">71</span> <span class="s1">'optimizer.optimizer'</span><span class="p">:</span> <span class="s1">'Adam'</span><span class="p">,</span>
|
||||
<span class="lineno">72</span> <span class="s1">'optimizer.learning_rate'</span><span class="p">:</span> <span class="mf">2.5e-4</span><span class="p">,</span>
|
||||
<span class="lineno">73</span> <span class="p">})</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-17'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-17'>#</a>
|
||||
</div>
|
||||
<p>设置保存/加载的模型</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">75</span> <span class="n">experiment</span><span class="o">.</span><span class="n">add_pytorch_models</span><span class="p">({</span><span class="s1">'model'</span><span class="p">:</span> <span class="n">conf</span><span class="o">.</span><span class="n">model</span><span class="p">})</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-18'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-18'>#</a>
|
||||
</div>
|
||||
<p>打印模型中参数的数量</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">77</span> <span class="n">logger</span><span class="o">.</span><span class="n">inspect</span><span class="p">(</span><span class="n">params</span><span class="o">=</span><span class="p">(</span><span class="nb">sum</span><span class="p">(</span><span class="n">p</span><span class="o">.</span><span class="n">numel</span><span class="p">()</span> <span class="k">for</span> <span class="n">p</span> <span class="ow">in</span> <span class="n">conf</span><span class="o">.</span><span class="n">model</span><span class="o">.</span><span class="n">parameters</span><span class="p">()</span> <span class="k">if</span> <span class="n">p</span><span class="o">.</span><span class="n">requires_grad</span><span class="p">)))</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-19'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-19'>#</a>
|
||||
</div>
|
||||
<p>开始实验并运行训练循环</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">79</span> <span class="k">with</span> <span class="n">experiment</span><span class="o">.</span><span class="n">start</span><span class="p">():</span>
|
||||
<span class="lineno">80</span> <span class="n">conf</span><span class="o">.</span><span class="n">run</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-20'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-20'>#</a>
|
||||
</div>
|
||||
<p></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">84</span><span class="k">if</span> <span class="vm">__name__</span> <span class="o">==</span> <span class="s1">'__main__'</span><span class="p">:</span>
|
||||
<span class="lineno">85</span> <span class="n">main</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
Binary file not shown.
|
After Width: | Height: | Size: 814 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 771 KiB |
@@ -0,0 +1,902 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content="这会产生算术问题。"/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="算术数据集"/>
|
||||
<meta name="twitter:description" content="这会产生算术问题。"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/experiments/arithmetic_dataset.html"/>
|
||||
<meta property="og:title" content="算术数据集"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="算术数据集"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="算术数据集"/>
|
||||
<meta property="og:description" content="这会产生算术问题。"/>
|
||||
|
||||
<title>算术数据集</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/experiments/arithmetic_dataset.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="index.html">experiments</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/experiments/arithmetic_dataset.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<p><em>这是基于<a href="https://twitter.com/gharik">乔治·哈里克（@gharik）</a>的代码。</em></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">11</span><span></span><span class="kn">import</span> <span class="nn">random</span>
|
||||
<span class="lineno">12</span><span class="kn">import</span> <span class="nn">string</span>
|
||||
<span class="lineno">13</span><span class="kn">from</span> <span class="nn">typing</span> <span class="kn">import</span> <span class="n">List</span>
|
||||
<span class="lineno">14</span>
|
||||
<span class="lineno">15</span><span class="kn">import</span> <span class="nn">torch</span>
|
||||
<span class="lineno">16</span><span class="kn">from</span> <span class="nn">labml.logger</span> <span class="kn">import</span> <span class="n">Text</span>
|
||||
<span class="lineno">17</span><span class="kn">from</span> <span class="nn">torch.utils.data</span> <span class="kn">import</span> <span class="n">DataLoader</span><span class="p">,</span> <span class="n">Dataset</span>
|
||||
<span class="lineno">18</span>
|
||||
<span class="lineno">19</span><span class="kn">from</span> <span class="nn">labml</span> <span class="kn">import</span> <span class="n">monit</span><span class="p">,</span> <span class="n">logger</span><span class="p">,</span> <span class="n">tracker</span>
|
||||
<span class="lineno">20</span><span class="kn">from</span> <span class="nn">labml.configs</span> <span class="kn">import</span> <span class="n">option</span>
|
||||
<span class="lineno">21</span><span class="kn">from</span> <span class="nn">labml_nn.experiments.nlp_autoregression</span> <span class="kn">import</span> <span class="n">NLPAutoRegressionConfigs</span><span class="p">,</span> <span class="n">transpose_batch</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-1'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-1'>#</a>
|
||||
</div>
|
||||
<h2>算术数据集</h2>
|
||||
<p>这会产生算术加法问题和运作解决方案。到目前为止，我们只实施了加法。</p>
|
||||
<p>它基于角色级别的标记化。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">24</span><span class="k">class</span> <span class="nc">ArithmeticDataset</span><span class="p">(</span><span class="n">Dataset</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-2'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-2'>#</a>
|
||||
</div>
|
||||
<ul><li><code class="highlight"><span></span><span class="n">seq_len</span></code>
|
||||
是生成的数学问题的序列长度。我们尽可能多地填写问题，直到这个长度：max_digits: 是操作数中的最大位数整数:n_sequences: 是每个纪元的序列数</li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">34</span> <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">seq_len</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">max_digits</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">n_sequences</span><span class="p">:</span> <span class="nb">int</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-3'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-3'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">41</span> <span class="bp">self</span><span class="o">.</span><span class="n">n_sequences</span> <span class="o">=</span> <span class="n">n_sequences</span>
|
||||
<span class="lineno">42</span> <span class="bp">self</span><span class="o">.</span><span class="n">max_digits</span> <span class="o">=</span> <span class="n">max_digits</span>
|
||||
<span class="lineno">43</span> <span class="bp">self</span><span class="o">.</span><span class="n">seq_len</span> <span class="o">=</span> <span class="n">seq_len</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-4'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-4'>#</a>
|
||||
</div>
|
||||
<p>令牌 ID 转换为字符串</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">45</span> <span class="bp">self</span><span class="o">.</span><span class="n">itos</span> <span class="o">=</span> <span class="nb">list</span><span class="p">(</span><span class="n">string</span><span class="o">.</span><span class="n">digits</span> <span class="o">+</span> <span class="s1">'xe =</span><span class="se">\n</span><span class="s1">?+;'</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-5'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-5'>#</a>
|
||||
</div>
|
||||
<p>字符到令牌 ID</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">47</span> <span class="bp">self</span><span class="o">.</span><span class="n">stoi</span> <span class="o">=</span> <span class="p">{</span><span class="n">c</span><span class="p">:</span> <span class="n">i</span> <span class="k">for</span> <span class="n">i</span><span class="p">,</span> <span class="n">c</span> <span class="ow">in</span> <span class="nb">enumerate</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">itos</span><span class="p">)}</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-6'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-6'>#</a>
|
||||
</div>
|
||||
<p>生成一个包含位<code class="highlight"><span></span><span class="n">n_digit</span></code>
|
||||
数的整数</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">49</span> <span class="nd">@staticmethod</span>
|
||||
<span class="lineno">50</span> <span class="k">def</span> <span class="nf">make_int</span><span class="p">(</span><span class="n">n_digits</span><span class="p">:</span> <span class="nb">int</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-7'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-7'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">54</span> <span class="n">res</span> <span class="o">=</span> <span class="mi">0</span>
|
||||
<span class="lineno">55</span> <span class="k">for</span> <span class="n">i</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="n">n_digits</span><span class="p">):</span>
|
||||
<span class="lineno">56</span> <span class="n">d</span> <span class="o">=</span> <span class="n">random</span><span class="o">.</span><span class="n">randrange</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="mi">11</span><span class="p">)</span> <span class="k">if</span> <span class="n">i</span> <span class="o">==</span> <span class="mi">0</span> <span class="k">else</span> <span class="n">random</span><span class="o">.</span><span class="n">randrange</span><span class="p">(</span><span class="mi">0</span><span class="p">,</span> <span class="mi">11</span><span class="p">)</span>
|
||||
<span class="lineno">57</span> <span class="n">res</span> <span class="o">=</span> <span class="n">res</span> <span class="o">*</span> <span class="mi">10</span> <span class="o">+</span> <span class="n">d</span>
|
||||
<span class="lineno">58</span>
|
||||
<span class="lineno">59</span> <span class="k">return</span> <span class="n">res</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-8'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-8'>#</a>
|
||||
</div>
|
||||
<p>生成的工作原理<code class="highlight"><span></span><span class="n">x</span> <span class="o">+</span> <span class="n">y</span></code>
|
||||
。例如，<code class="highlight"><span></span><span class="mi">11</span><span class="o">+</span><span class="mi">29</span></code>
|
||||
它会生成<code class="highlight"><span></span><span class="mf">1e0</span><span class="o">+</span><span class="mf">9e0</span><span class="o">+</span><span class="mf">0e0</span><span class="o">=</span><span class="mf">10e0</span> <span class="mf">1e0</span><span class="o">+</span><span class="mf">2e0</span><span class="o">+</span><span class="mf">1e0</span><span class="o">=</span><span class="mf">4e0</span></code>
|
||||
。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">61</span> <span class="nd">@staticmethod</span>
|
||||
<span class="lineno">62</span> <span class="k">def</span> <span class="nf">get_add_explanation</span><span class="p">(</span><span class="n">x</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">y</span><span class="p">:</span> <span class="nb">int</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-9'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-9'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">69</span> <span class="n">carry</span> <span class="o">=</span> <span class="mi">0</span>
|
||||
<span class="lineno">70</span> <span class="n">e</span> <span class="o">=</span> <span class="mi">0</span>
|
||||
<span class="lineno">71</span> <span class="n">explanation</span> <span class="o">=</span> <span class="p">[]</span>
|
||||
<span class="lineno">72</span> <span class="k">while</span> <span class="n">x</span> <span class="o">></span> <span class="mi">0</span> <span class="ow">or</span> <span class="n">y</span> <span class="o">></span> <span class="mi">0</span> <span class="ow">or</span> <span class="n">carry</span> <span class="o">></span> <span class="mi">0</span><span class="p">:</span>
|
||||
<span class="lineno">73</span> <span class="n">rx</span><span class="p">,</span> <span class="n">ry</span> <span class="o">=</span> <span class="n">x</span> <span class="o">%</span> <span class="mi">10</span><span class="p">,</span> <span class="n">y</span> <span class="o">%</span> <span class="mi">10</span>
|
||||
<span class="lineno">74</span> <span class="n">total</span> <span class="o">=</span> <span class="n">rx</span> <span class="o">+</span> <span class="n">ry</span> <span class="o">+</span> <span class="n">carry</span>
|
||||
<span class="lineno">75</span> <span class="n">explanation</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="sa">f</span><span class="s2">"</span><span class="si">{</span><span class="n">rx</span><span class="si">}</span><span class="s2">e</span><span class="si">{</span><span class="n">e</span><span class="si">}</span><span class="s2">+</span><span class="si">{</span><span class="n">ry</span><span class="si">}</span><span class="s2">e</span><span class="si">{</span><span class="n">e</span><span class="si">}</span><span class="s2">+</span><span class="si">{</span><span class="n">carry</span><span class="si">}</span><span class="s2">e</span><span class="si">{</span><span class="n">e</span><span class="si">}</span><span class="s2">==</span><span class="si">{</span><span class="n">total</span><span class="si">}</span><span class="s2">e</span><span class="si">{</span><span class="n">e</span><span class="si">}</span><span class="s2">"</span><span class="p">)</span>
|
||||
<span class="lineno">76</span> <span class="n">x</span><span class="p">,</span> <span class="n">y</span><span class="p">,</span> <span class="n">carry</span> <span class="o">=</span> <span class="n">x</span> <span class="o">//</span> <span class="mi">10</span><span class="p">,</span> <span class="n">y</span> <span class="o">//</span> <span class="mi">10</span><span class="p">,</span> <span class="n">total</span> <span class="o">//</span> <span class="mi">10</span>
|
||||
<span class="lineno">77</span> <span class="n">e</span> <span class="o">+=</span> <span class="mi">1</span>
|
||||
<span class="lineno">78</span>
|
||||
<span class="lineno">79</span> <span class="k">return</span> <span class="s1">' '</span><span class="o">.</span><span class="n">join</span><span class="p">(</span><span class="n">explanation</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-10'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-10'>#</a>
|
||||
</div>
|
||||
<p>不管是否用 pre_explansion 问问题</p>
|
||||
<p>用运作和答案创建算术加法问题。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">82</span> <span class="k">def</span> <span class="nf">make_add_problem</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-11'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-11'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">86</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">make_int</span><span class="p">(</span><span class="n">n_digits</span><span class="o">=</span><span class="n">random</span><span class="o">.</span><span class="n">randrange</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">max_digits</span> <span class="o">+</span> <span class="mi">1</span><span class="p">))</span>
|
||||
<span class="lineno">87</span> <span class="n">y</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">make_int</span><span class="p">(</span><span class="n">n_digits</span><span class="o">=</span><span class="n">random</span><span class="o">.</span><span class="n">randrange</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">max_digits</span> <span class="o">+</span> <span class="mi">1</span><span class="p">))</span>
|
||||
<span class="lineno">88</span>
|
||||
<span class="lineno">89</span> <span class="n">explanation</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">get_add_explanation</span><span class="p">(</span><span class="n">x</span><span class="p">,</span> <span class="n">y</span><span class="p">)</span>
|
||||
<span class="lineno">90</span> <span class="k">return</span> <span class="sa">f</span><span class="s2">"x=</span><span class="si">{</span><span class="n">x</span><span class="si">}</span><span class="s2">+</span><span class="si">{</span><span class="n">y</span><span class="si">}</span><span class="s2">; </span><span class="si">{</span><span class="n">explanation</span><span class="si">}</span><span class="s2"> x==</span><span class="si">{</span><span class="n">x</span><span class="w"> </span><span class="o">+</span><span class="w"> </span><span class="n">y</span><span class="si">}</span><span class="se">\n</span><span class="s2">"</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-12'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-12'>#</a>
|
||||
</div>
|
||||
<p>获取算术问题和答案。这用于评估。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">92</span> <span class="k">def</span> <span class="nf">get_qa</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-13'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-13'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">96</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">make_int</span><span class="p">(</span><span class="n">n_digits</span><span class="o">=</span><span class="n">random</span><span class="o">.</span><span class="n">randrange</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">max_digits</span> <span class="o">+</span> <span class="mi">1</span><span class="p">))</span>
|
||||
<span class="lineno">97</span> <span class="n">y</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">make_int</span><span class="p">(</span><span class="n">n_digits</span><span class="o">=</span><span class="n">random</span><span class="o">.</span><span class="n">randrange</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">max_digits</span> <span class="o">+</span> <span class="mi">1</span><span class="p">))</span>
|
||||
<span class="lineno">98</span>
|
||||
<span class="lineno">99</span> <span class="k">return</span> <span class="sa">f</span><span class="s1">'x=</span><span class="si">{</span><span class="n">x</span><span class="si">}</span><span class="s1">+</span><span class="si">{</span><span class="n">y</span><span class="si">}</span><span class="s1">;'</span><span class="p">,</span> <span class="sa">f</span><span class="s1">'</span><span class="si">{</span><span class="n">x</span><span class="w"> </span><span class="o">+</span><span class="w"> </span><span class="n">y</span><span class="si">}</span><span class="s1">'</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-14'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-14'>#</a>
|
||||
</div>
|
||||
<p>生成多个问题并将它们打包成一个序列。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">101</span> <span class="k">def</span> <span class="nf">get_packed_math_input</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-15'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-15'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">105</span> <span class="n">s_enc</span> <span class="o">=</span> <span class="p">[]</span>
|
||||
<span class="lineno">106</span> <span class="k">while</span> <span class="nb">len</span><span class="p">(</span><span class="n">s_enc</span><span class="p">)</span> <span class="o"><=</span> <span class="bp">self</span><span class="o">.</span><span class="n">seq_len</span><span class="p">:</span>
|
||||
<span class="lineno">107</span> <span class="n">s_part</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">make_add_problem</span><span class="p">()</span>
|
||||
<span class="lineno">108</span> <span class="n">s_part_enc</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">encode</span><span class="p">(</span><span class="s1">'?'</span> <span class="o">+</span> <span class="n">s_part</span><span class="p">)</span>
|
||||
<span class="lineno">109</span> <span class="n">s_enc</span> <span class="o">=</span> <span class="n">s_enc</span> <span class="o">+</span> <span class="n">s_part_enc</span>
|
||||
<span class="lineno">110</span> <span class="k">return</span> <span class="n">s_enc</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-16'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-16'>#</a>
|
||||
</div>
|
||||
<p>对给定字符串进行编码</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">112</span> <span class="k">def</span> <span class="nf">encode</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">s</span><span class="p">:</span> <span class="nb">str</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-17'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-17'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">116</span> <span class="k">return</span> <span class="p">[</span><span class="bp">self</span><span class="o">.</span><span class="n">stoi</span><span class="p">[</span><span class="n">c</span><span class="p">]</span> <span class="k">for</span> <span class="n">c</span> <span class="ow">in</span> <span class="n">s</span><span class="p">]</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-18'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-18'>#</a>
|
||||
</div>
|
||||
<p>解码令牌 ID 列表</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">118</span> <span class="k">def</span> <span class="nf">decode</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">arr</span><span class="p">:</span> <span class="n">List</span><span class="p">[</span><span class="nb">int</span><span class="p">]):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-19'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-19'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">122</span> <span class="k">return</span> <span class="s1">''</span><span class="o">.</span><span class="n">join</span><span class="p">([</span><span class="bp">self</span><span class="o">.</span><span class="n">itos</span><span class="p">[</span><span class="n">c</span><span class="p">]</span> <span class="k">for</span> <span class="n">c</span> <span class="ow">in</span> <span class="n">arr</span><span class="p">])</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-20'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-20'>#</a>
|
||||
</div>
|
||||
<p>获取自动回归建模的输入和目标对</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">124</span> <span class="k">def</span> <span class="fm">__getitem__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">idx</span><span class="p">:</span> <span class="nb">int</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-21'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-21'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">128</span> <span class="n">s</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">tensor</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">get_packed_math_input</span><span class="p">())</span>
|
||||
<span class="lineno">129</span> <span class="k">return</span> <span class="n">s</span><span class="p">[:</span><span class="bp">self</span><span class="o">.</span><span class="n">seq_len</span><span class="p">],</span> <span class="n">s</span><span class="p">[</span><span class="mi">1</span><span class="p">:</span><span class="bp">self</span><span class="o">.</span><span class="n">seq_len</span> <span class="o">+</span> <span class="mi">1</span><span class="p">]</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-22'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-22'>#</a>
|
||||
</div>
|
||||
<p>每个纪元的序列数</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">131</span> <span class="k">def</span> <span class="fm">__len__</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-23'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-23'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">135</span> <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">n_sequences</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-24'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-24'>#</a>
|
||||
</div>
|
||||
<h2>算术任务实验配置</h2>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">138</span><span class="k">class</span> <span class="nc">ArithmeticAutoregression</span><span class="p">(</span><span class="n">NLPAutoRegressionConfigs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-25'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-25'>#</a>
|
||||
</div>
|
||||
<p>每个操作数整数的最大位数</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">143</span> <span class="n">max_digits</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">4</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-26'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-26'>#</a>
|
||||
</div>
|
||||
<p>每个纪元的训练序列数</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">145</span> <span class="n">train_sequences_per_epoch</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">2</span> <span class="o">**</span> <span class="mi">12</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-27'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-27'>#</a>
|
||||
</div>
|
||||
<p>训练数据加载器</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">147</span> <span class="n">train_loader</span><span class="p">:</span> <span class="n">DataLoader</span> <span class="o">=</span> <span class="s1">'arithmetic_train_loader'</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-28'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-28'>#</a>
|
||||
</div>
|
||||
<p>评估中的问题数量</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">149</span> <span class="n">n_tests</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">64</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-29'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-29'>#</a>
|
||||
</div>
|
||||
<p>不需要验证数据集</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">151</span> <span class="n">validator</span> <span class="o">=</span> <span class="kc">None</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-30'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-30'>#</a>
|
||||
</div>
|
||||
<p>每个纪元运行评估的次数</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">153</span> <span class="n">inner_iterations</span> <span class="o">=</span> <span class="mi">4</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-31'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-31'>#</a>
|
||||
</div>
|
||||
<p>词汇表中的代币数量</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">155</span> <span class="n">n_tokens</span> <span class="o">=</span> <span class="nb">len</span><span class="p">(</span><span class="n">ArithmeticDataset</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span><span class="o">.</span><span class="n">itos</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-32'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-32'>#</a>
|
||||
</div>
|
||||
<h3>评估</h3>
|
||||
<p>我们使用采样函数来评估一组问题的模型</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">157</span> <span class="nd">@torch</span><span class="o">.</span><span class="n">no_grad</span><span class="p">()</span>
|
||||
<span class="lineno">158</span> <span class="k">def</span> <span class="nf">sample</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-33'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-33'>#</a>
|
||||
</div>
|
||||
<p>跳过第一个纪元</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">166</span> <span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">training_loop</span><span class="o">.</span><span class="n">idx</span> <span class="o"><</span> <span class="mi">1</span><span class="p">:</span>
|
||||
<span class="lineno">167</span> <span class="k">return</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-34'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-34'>#</a>
|
||||
</div>
|
||||
<p>创建数据集以生成问题</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">170</span> <span class="n">dataset</span> <span class="o">=</span> <span class="n">ArithmeticDataset</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">seq_len</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">max_digits</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-35'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-35'>#</a>
|
||||
</div>
|
||||
<p>获取一系列问题和答案</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">172</span> <span class="n">qa</span> <span class="o">=</span> <span class="p">[</span><span class="n">dataset</span><span class="o">.</span><span class="n">get_qa</span><span class="p">()</span> <span class="k">for</span> <span class="n">_</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">n_tests</span><span class="p">)]</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-36'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-36'>#</a>
|
||||
</div>
|
||||
<p>只收集问题</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">174</span> <span class="n">questions</span> <span class="o">=</span> <span class="p">[</span><span class="n">p</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span> <span class="k">for</span> <span class="n">p</span> <span class="ow">in</span> <span class="n">qa</span><span class="p">]</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-37'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-37'>#</a>
|
||||
</div>
|
||||
<p>仅使用初始令牌创建张量</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">177</span> <span class="n">data</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">tensor</span><span class="p">([[</span><span class="n">dataset</span><span class="o">.</span><span class="n">stoi</span><span class="p">[</span><span class="n">p</span><span class="p">[</span><span class="mi">0</span><span class="p">]]</span> <span class="k">for</span> <span class="n">p</span> <span class="ow">in</span> <span class="n">questions</span><span class="p">]])</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-38'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-38'>#</a>
|
||||
</div>
|
||||
<p>移至设备</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">179</span> <span class="n">data</span> <span class="o">=</span> <span class="n">data</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">device</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-39'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-39'>#</a>
|
||||
</div>
|
||||
<p>已完成的序列数</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">182</span> <span class="n">finished</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">zeros</span><span class="p">((</span><span class="nb">len</span><span class="p">(</span><span class="n">questions</span><span class="p">),))</span><span class="o">.</span><span class="n">bool</span><span class="p">()</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">device</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-40'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-40'>#</a>
|
||||
</div>
|
||||
<p>换行符的标记 ID-这标志着答案的结束</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">184</span> <span class="n">new_line</span> <span class="o">=</span> <span class="n">dataset</span><span class="o">.</span><span class="n">stoi</span><span class="p">[</span><span class="s1">'</span><span class="se">\n</span><span class="s1">'</span><span class="p">]</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-41'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-41'>#</a>
|
||||
</div>
|
||||
<p>抽样结果</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">187</span> <span class="n">results</span> <span class="o">=</span> <span class="p">[</span><span class="n">p</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span> <span class="k">for</span> <span class="n">p</span> <span class="ow">in</span> <span class="n">questions</span><span class="p">]</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-42'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-42'>#</a>
|
||||
</div>
|
||||
<p>样本直至序列长度</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">190</span> <span class="k">for</span> <span class="n">i</span> <span class="ow">in</span> <span class="n">monit</span><span class="o">.</span><span class="n">iterate</span><span class="p">(</span><span class="s1">'Sample'</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">seq_len</span> <span class="o">-</span> <span class="mi">1</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-43'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-43'>#</a>
|
||||
</div>
|
||||
<p>如果所有的序列都完成了，我们就跳过这个</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">192</span> <span class="k">if</span> <span class="n">finished</span><span class="o">.</span><span class="n">sum</span><span class="p">()</span> <span class="o">==</span> <span class="nb">len</span><span class="p">(</span><span class="n">finished</span><span class="p">):</span>
|
||||
<span class="lineno">193</span> <span class="k">continue</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-44'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-44'>#</a>
|
||||
</div>
|
||||
<p>获取模型输出</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">196</span> <span class="n">output</span><span class="p">,</span> <span class="o">*</span><span class="n">_</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">model</span><span class="p">(</span><span class="n">data</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-45'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-45'>#</a>
|
||||
</div>
|
||||
<p>获取模型预测（贪婪）</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">198</span> <span class="n">output</span> <span class="o">=</span> <span class="n">output</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">]</span><span class="o">.</span><span class="n">argmax</span><span class="p">(</span><span class="n">dim</span><span class="o">=-</span><span class="mi">1</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-46'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-46'>#</a>
|
||||
</div>
|
||||
<p>找出哪些序列已完成</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">201</span> <span class="n">finished</span> <span class="o">=</span> <span class="n">finished</span> <span class="o">|</span> <span class="p">(</span><span class="n">output</span> <span class="o">==</span> <span class="n">new_line</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-47'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-47'>#</a>
|
||||
</div>
|
||||
<p>如果全部完成，则跳过</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">203</span> <span class="k">if</span> <span class="n">finished</span><span class="o">.</span><span class="n">sum</span><span class="p">()</span> <span class="o">==</span> <span class="nb">len</span><span class="p">(</span><span class="n">finished</span><span class="p">):</span>
|
||||
<span class="lineno">204</span> <span class="k">continue</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-48'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-48'>#</a>
|
||||
</div>
|
||||
<p>用问题覆盖</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">207</span> <span class="k">for</span> <span class="n">j</span><span class="p">,</span> <span class="n">p</span> <span class="ow">in</span> <span class="nb">enumerate</span><span class="p">(</span><span class="n">questions</span><span class="p">):</span>
|
||||
<span class="lineno">208</span> <span class="k">if</span> <span class="nb">len</span><span class="p">(</span><span class="n">p</span><span class="p">)</span> <span class="o">></span> <span class="n">i</span> <span class="o">+</span> <span class="mi">1</span><span class="p">:</span>
|
||||
<span class="lineno">209</span> <span class="n">output</span><span class="p">[</span><span class="n">j</span><span class="p">]</span> <span class="o">=</span> <span class="n">dataset</span><span class="o">.</span><span class="n">stoi</span><span class="p">[</span><span class="n">p</span><span class="p">[</span><span class="n">i</span> <span class="o">+</span> <span class="mi">1</span><span class="p">]]</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-49'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-49'>#</a>
|
||||
</div>
|
||||
<p>将下一个令牌添加到输入中</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">212</span> <span class="n">data</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">cat</span><span class="p">([</span><span class="n">data</span><span class="p">,</span> <span class="n">output</span><span class="p">[</span><span class="kc">None</span><span class="p">,</span> <span class="p">:]],</span> <span class="n">dim</span><span class="o">=</span><span class="mi">0</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-50'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-50'>#</a>
|
||||
</div>
|
||||
<p>获取抽样结果</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">215</span> <span class="k">for</span> <span class="n">j</span><span class="p">,</span> <span class="n">c</span> <span class="ow">in</span> <span class="nb">enumerate</span><span class="p">(</span><span class="n">output</span><span class="p">):</span>
|
||||
<span class="lineno">216</span> <span class="n">results</span><span class="p">[</span><span class="n">j</span><span class="p">]</span> <span class="o">+=</span> <span class="n">dataset</span><span class="o">.</span><span class="n">itos</span><span class="p">[</span><span class="n">c</span><span class="p">]</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-51'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-51'>#</a>
|
||||
</div>
|
||||
<p>丢弃结果中答案后的所有内容</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">219</span> <span class="n">results</span> <span class="o">=</span> <span class="p">[</span><span class="n">r</span><span class="o">.</span><span class="n">split</span><span class="p">(</span><span class="s1">'</span><span class="se">\n</span><span class="s1">'</span><span class="p">)[</span><span class="mi">0</span><span class="p">]</span> <span class="k">for</span> <span class="n">r</span> <span class="ow">in</span> <span class="n">results</span><span class="p">]</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-52'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-52'>#</a>
|
||||
</div>
|
||||
<p>记录样本</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">222</span> <span class="n">res_sample</span> <span class="o">=</span> <span class="n">results</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span><span class="o">.</span><span class="n">split</span><span class="p">(</span><span class="s1">';'</span><span class="p">)</span>
|
||||
<span class="lineno">223</span> <span class="n">logger</span><span class="o">.</span><span class="n">log</span><span class="p">([(</span><span class="n">res_sample</span><span class="p">[</span><span class="mi">0</span><span class="p">],</span> <span class="n">Text</span><span class="o">.</span><span class="n">key</span><span class="p">),</span> <span class="p">(</span><span class="s1">';'</span><span class="p">,</span> <span class="n">Text</span><span class="o">.</span><span class="n">subtle</span><span class="p">),</span> <span class="p">(</span><span class="s1">';'</span><span class="o">.</span><span class="n">join</span><span class="p">(</span><span class="n">res_sample</span><span class="p">[</span><span class="mi">1</span><span class="p">:]),</span> <span class="n">Text</span><span class="o">.</span><span class="n">none</span><span class="p">)])</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-53'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-53'>#</a>
|
||||
</div>
|
||||
<p>得到答案</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">226</span> <span class="n">results</span> <span class="o">=</span> <span class="p">[</span><span class="n">r</span><span class="o">.</span><span class="n">split</span><span class="p">(</span><span class="s1">'x=='</span><span class="p">)[</span><span class="o">-</span><span class="mi">1</span><span class="p">]</span> <span class="k">for</span> <span class="n">r</span> <span class="ow">in</span> <span class="n">results</span><span class="p">]</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-54'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-54'>#</a>
|
||||
</div>
|
||||
<p>计算正确答案的数量</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">229</span> <span class="n">correct</span> <span class="o">=</span> <span class="mi">0</span>
|
||||
<span class="lineno">230</span> <span class="k">for</span> <span class="n">r</span><span class="p">,</span> <span class="n">_qa</span> <span class="ow">in</span> <span class="nb">zip</span><span class="p">(</span><span class="n">results</span><span class="p">,</span> <span class="n">qa</span><span class="p">):</span>
|
||||
<span class="lineno">231</span> <span class="k">if</span> <span class="n">r</span> <span class="o">==</span> <span class="n">_qa</span><span class="p">[</span><span class="mi">1</span><span class="p">]:</span>
|
||||
<span class="lineno">232</span> <span class="n">correct</span> <span class="o">+=</span> <span class="mi">1</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-55'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-55'>#</a>
|
||||
</div>
|
||||
<p>记录分数</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">235</span> <span class="n">tracker</span><span class="o">.</span><span class="n">save</span><span class="p">(</span><span class="s1">'score'</span><span class="p">,</span> <span class="n">correct</span> <span class="o">/</span> <span class="nb">len</span><span class="p">(</span><span class="n">results</span><span class="p">))</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-56'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-56'>#</a>
|
||||
</div>
|
||||
<p>训练数据加载器</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">238</span><span class="nd">@option</span><span class="p">(</span><span class="n">ArithmeticAutoregression</span><span class="o">.</span><span class="n">train_loader</span><span class="p">)</span>
|
||||
<span class="lineno">239</span><span class="k">def</span> <span class="nf">arithmetic_train_loader</span><span class="p">(</span><span class="n">c</span><span class="p">:</span> <span class="n">ArithmeticAutoregression</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-57'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-57'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">243</span> <span class="k">return</span> <span class="n">DataLoader</span><span class="p">(</span><span class="n">ArithmeticDataset</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">seq_len</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">max_digits</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">train_sequences_per_epoch</span><span class="p">),</span>
|
||||
<span class="lineno">244</span> <span class="n">batch_size</span><span class="o">=</span><span class="n">c</span><span class="o">.</span><span class="n">batch_size</span><span class="p">,</span>
|
||||
<span class="lineno">245</span> <span class="n">collate_fn</span><span class="o">=</span><span class="n">transpose_batch</span><span class="p">,</span>
|
||||
<span class="lineno">246</span> <span class="n">num_workers</span><span class="o">=</span><span class="mi">4</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-58'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-58'>#</a>
|
||||
</div>
|
||||
<p>用于测试生成的问题的代码</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">249</span><span class="k">def</span> <span class="nf">_test</span><span class="p">():</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-59'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-59'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">253</span> <span class="n">dataset</span> <span class="o">=</span> <span class="n">ArithmeticDataset</span><span class="p">(</span><span class="mi">256</span><span class="p">,</span> <span class="mi">8</span><span class="p">,</span> <span class="mi">10</span><span class="p">)</span>
|
||||
<span class="lineno">254</span>
|
||||
<span class="lineno">255</span> <span class="nb">print</span><span class="p">(</span><span class="n">dataset</span><span class="o">.</span><span class="n">decode</span><span class="p">(</span><span class="n">dataset</span><span class="o">.</span><span class="n">get_packed_math_input</span><span class="p">()))</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-60'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-60'>#</a>
|
||||
</div>
|
||||
<p></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">259</span><span class="k">if</span> <span class="vm">__name__</span> <span class="o">==</span> <span class="s1">'__main__'</span><span class="p">:</span>
|
||||
<span class="lineno">260</span> <span class="n">_test</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
@@ -0,0 +1,443 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content="这是 CIFAR10 数据集的可重复使用的训练器"/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="CIFAR10 实验"/>
|
||||
<meta name="twitter:description" content="这是 CIFAR10 数据集的可重复使用的训练器"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/experiments/cifar10.html"/>
|
||||
<meta property="og:title" content="CIFAR10 实验"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="CIFAR10 实验"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="CIFAR10 实验"/>
|
||||
<meta property="og:description" content="这是 CIFAR10 数据集的可重复使用的训练器"/>
|
||||
|
||||
<title>CIFAR10 实验</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/experiments/cifar10.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="index.html">experiments</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/experiments/cifar10.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1>CIFAR10 实验</h1>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">10</span><span></span><span class="kn">from</span> <span class="nn">typing</span> <span class="kn">import</span> <span class="n">List</span>
|
||||
<span class="lineno">11</span>
|
||||
<span class="lineno">12</span><span class="kn">import</span> <span class="nn">torch.nn</span> <span class="k">as</span> <span class="nn">nn</span>
|
||||
<span class="lineno">13</span>
|
||||
<span class="lineno">14</span><span class="kn">from</span> <span class="nn">labml</span> <span class="kn">import</span> <span class="n">lab</span>
|
||||
<span class="lineno">15</span><span class="kn">from</span> <span class="nn">labml.configs</span> <span class="kn">import</span> <span class="n">option</span>
|
||||
<span class="lineno">16</span><span class="kn">from</span> <span class="nn">labml_helpers.datasets.cifar10</span> <span class="kn">import</span> <span class="n">CIFAR10Configs</span> <span class="k">as</span> <span class="n">CIFAR10DatasetConfigs</span>
|
||||
<span class="lineno">17</span><span class="kn">from</span> <span class="nn">labml_helpers.module</span> <span class="kn">import</span> <span class="n">Module</span>
|
||||
<span class="lineno">18</span><span class="kn">from</span> <span class="nn">labml_nn.experiments.mnist</span> <span class="kn">import</span> <span class="n">MNISTConfigs</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-1'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-1'>#</a>
|
||||
</div>
|
||||
<h2>配置</h2>
|
||||
<p>这是从和开始的 CIFAR 10 数据集配置扩展<a href="https://github.com/labmlai/labml/tree/master/helpers"><code class="highlight"><span></span><span class="n">labml_helpers</span></code>
|
||||
</a>而来的<a href="mnist.html"><code class="highlight"><span></span><span class="n">MNISTConfigs</span></code>
|
||||
</a>。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">21</span><span class="k">class</span> <span class="nc">CIFAR10Configs</span><span class="p">(</span><span class="n">CIFAR10DatasetConfigs</span><span class="p">,</span> <span class="n">MNISTConfigs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-2'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-2'>#</a>
|
||||
</div>
|
||||
<p>默认使用 CIFAR10 数据集</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">30</span> <span class="n">dataset_name</span><span class="p">:</span> <span class="nb">str</span> <span class="o">=</span> <span class="s1">'CIFAR10'</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-3'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-3'>#</a>
|
||||
</div>
|
||||
<h3>增强的 CIFAR 10 训练数据集</h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">33</span><span class="nd">@option</span><span class="p">(</span><span class="n">CIFAR10Configs</span><span class="o">.</span><span class="n">train_dataset</span><span class="p">)</span>
|
||||
<span class="lineno">34</span><span class="k">def</span> <span class="nf">cifar10_train_augmented</span><span class="p">():</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-4'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-4'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">38</span> <span class="kn">from</span> <span class="nn">torchvision.datasets</span> <span class="kn">import</span> <span class="n">CIFAR10</span>
|
||||
<span class="lineno">39</span> <span class="kn">from</span> <span class="nn">torchvision.transforms</span> <span class="kn">import</span> <span class="n">transforms</span>
|
||||
<span class="lineno">40</span> <span class="k">return</span> <span class="n">CIFAR10</span><span class="p">(</span><span class="nb">str</span><span class="p">(</span><span class="n">lab</span><span class="o">.</span><span class="n">get_data_path</span><span class="p">()),</span>
|
||||
<span class="lineno">41</span> <span class="n">train</span><span class="o">=</span><span class="kc">True</span><span class="p">,</span>
|
||||
<span class="lineno">42</span> <span class="n">download</span><span class="o">=</span><span class="kc">True</span><span class="p">,</span>
|
||||
<span class="lineno">43</span> <span class="n">transform</span><span class="o">=</span><span class="n">transforms</span><span class="o">.</span><span class="n">Compose</span><span class="p">([</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-5'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-5'>#</a>
|
||||
</div>
|
||||
<p>填充和裁剪</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">45</span> <span class="n">transforms</span><span class="o">.</span><span class="n">RandomCrop</span><span class="p">(</span><span class="mi">32</span><span class="p">,</span> <span class="n">padding</span><span class="o">=</span><span class="mi">4</span><span class="p">),</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-6'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-6'>#</a>
|
||||
</div>
|
||||
<p>随机水平翻转</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">47</span> <span class="n">transforms</span><span class="o">.</span><span class="n">RandomHorizontalFlip</span><span class="p">(),</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-7'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-7'>#</a>
|
||||
</div>
|
||||
<p></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">49</span> <span class="n">transforms</span><span class="o">.</span><span class="n">ToTensor</span><span class="p">(),</span>
|
||||
<span class="lineno">50</span> <span class="n">transforms</span><span class="o">.</span><span class="n">Normalize</span><span class="p">((</span><span class="mf">0.5</span><span class="p">,</span> <span class="mf">0.5</span><span class="p">,</span> <span class="mf">0.5</span><span class="p">),</span> <span class="p">(</span><span class="mf">0.5</span><span class="p">,</span> <span class="mf">0.5</span><span class="p">,</span> <span class="mf">0.5</span><span class="p">))</span>
|
||||
<span class="lineno">51</span> <span class="p">]))</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-8'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-8'>#</a>
|
||||
</div>
|
||||
<h3>非增强 CIFAR 10 验证数据集</h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">54</span><span class="nd">@option</span><span class="p">(</span><span class="n">CIFAR10Configs</span><span class="o">.</span><span class="n">valid_dataset</span><span class="p">)</span>
|
||||
<span class="lineno">55</span><span class="k">def</span> <span class="nf">cifar10_valid_no_augment</span><span class="p">():</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-9'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-9'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">59</span> <span class="kn">from</span> <span class="nn">torchvision.datasets</span> <span class="kn">import</span> <span class="n">CIFAR10</span>
|
||||
<span class="lineno">60</span> <span class="kn">from</span> <span class="nn">torchvision.transforms</span> <span class="kn">import</span> <span class="n">transforms</span>
|
||||
<span class="lineno">61</span> <span class="k">return</span> <span class="n">CIFAR10</span><span class="p">(</span><span class="nb">str</span><span class="p">(</span><span class="n">lab</span><span class="o">.</span><span class="n">get_data_path</span><span class="p">()),</span>
|
||||
<span class="lineno">62</span> <span class="n">train</span><span class="o">=</span><span class="kc">False</span><span class="p">,</span>
|
||||
<span class="lineno">63</span> <span class="n">download</span><span class="o">=</span><span class="kc">True</span><span class="p">,</span>
|
||||
<span class="lineno">64</span> <span class="n">transform</span><span class="o">=</span><span class="n">transforms</span><span class="o">.</span><span class="n">Compose</span><span class="p">([</span>
|
||||
<span class="lineno">65</span> <span class="n">transforms</span><span class="o">.</span><span class="n">ToTensor</span><span class="p">(),</span>
|
||||
<span class="lineno">66</span> <span class="n">transforms</span><span class="o">.</span><span class="n">Normalize</span><span class="p">((</span><span class="mf">0.5</span><span class="p">,</span> <span class="mf">0.5</span><span class="p">,</span> <span class="mf">0.5</span><span class="p">),</span> <span class="p">(</span><span class="mf">0.5</span><span class="p">,</span> <span class="mf">0.5</span><span class="p">,</span> <span class="mf">0.5</span><span class="p">))</span>
|
||||
<span class="lineno">67</span> <span class="p">]))</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-10'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-10'>#</a>
|
||||
</div>
|
||||
<h3>用于 CIFAR-10 分类的 VGG 模型</h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">70</span><span class="k">class</span> <span class="nc">CIFAR10VGGModel</span><span class="p">(</span><span class="n">Module</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-11'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-11'>#</a>
|
||||
</div>
|
||||
<p>卷积和激活相结合</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">75</span> <span class="k">def</span> <span class="nf">conv_block</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">in_channels</span><span class="p">,</span> <span class="n">out_channels</span><span class="p">)</span> <span class="o">-></span> <span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">:</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-12'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-12'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">79</span> <span class="k">return</span> <span class="n">nn</span><span class="o">.</span><span class="n">Sequential</span><span class="p">(</span>
|
||||
<span class="lineno">80</span> <span class="n">nn</span><span class="o">.</span><span class="n">Conv2d</span><span class="p">(</span><span class="n">in_channels</span><span class="p">,</span> <span class="n">out_channels</span><span class="p">,</span> <span class="n">kernel_size</span><span class="o">=</span><span class="mi">3</span><span class="p">,</span> <span class="n">padding</span><span class="o">=</span><span class="mi">1</span><span class="p">),</span>
|
||||
<span class="lineno">81</span> <span class="n">nn</span><span class="o">.</span><span class="n">ReLU</span><span class="p">(</span><span class="n">inplace</span><span class="o">=</span><span class="kc">True</span><span class="p">),</span>
|
||||
<span class="lineno">82</span> <span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-13'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-13'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">84</span> <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">blocks</span><span class="p">:</span> <span class="n">List</span><span class="p">[</span><span class="n">List</span><span class="p">[</span><span class="nb">int</span><span class="p">]]):</span>
|
||||
<span class="lineno">85</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-14'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-14'>#</a>
|
||||
</div>
|
||||
<p>5 个<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.72777em;vertical-align:-0.08333em;"></span><span class="mord">2</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin">×</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span></span><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord">2</span></span></span></span></span>池化图层将生成大小为 size 的输出<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.65952em;vertical-align:0em;"></span><span class="mord">1</span><span class="mspace"> </span><span class="mord mathnormal">t</span><span class="mord mathnormal">im</span><span class="mord mathnormal">es</span><span class="mord">1</span></span></span></span></span>。CIFAR 10 图像大小为<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.72777em;vertical-align:-0.08333em;"></span><span class="mord">32</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin">×</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span></span><span class="base"><span class="strut" style="height:0.64444em;vertical-align:0em;"></span><span class="mord">32</span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">89</span> <span class="k">assert</span> <span class="nb">len</span><span class="p">(</span><span class="n">blocks</span><span class="p">)</span> <span class="o">==</span> <span class="mi">5</span>
|
||||
<span class="lineno">90</span> <span class="n">layers</span> <span class="o">=</span> <span class="p">[]</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-15'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-15'>#</a>
|
||||
</div>
|
||||
<p>RGB 通道</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">92</span> <span class="n">in_channels</span> <span class="o">=</span> <span class="mi">3</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-16'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-16'>#</a>
|
||||
</div>
|
||||
<p>每个区块中每层的通道数</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">94</span> <span class="k">for</span> <span class="n">block</span> <span class="ow">in</span> <span class="n">blocks</span><span class="p">:</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-17'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-17'>#</a>
|
||||
</div>
|
||||
<p>卷积、归一化和激活层</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">96</span> <span class="k">for</span> <span class="n">channels</span> <span class="ow">in</span> <span class="n">block</span><span class="p">:</span>
|
||||
<span class="lineno">97</span> <span class="n">layers</span> <span class="o">+=</span> <span class="bp">self</span><span class="o">.</span><span class="n">conv_block</span><span class="p">(</span><span class="n">in_channels</span><span class="p">,</span> <span class="n">channels</span><span class="p">)</span>
|
||||
<span class="lineno">98</span> <span class="n">in_channels</span> <span class="o">=</span> <span class="n">channels</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-18'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-18'>#</a>
|
||||
</div>
|
||||
<p>每个区块末端的最大池数</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">100</span> <span class="n">layers</span> <span class="o">+=</span> <span class="p">[</span><span class="n">nn</span><span class="o">.</span><span class="n">MaxPool2d</span><span class="p">(</span><span class="n">kernel_size</span><span class="o">=</span><span class="mi">2</span><span class="p">,</span> <span class="n">stride</span><span class="o">=</span><span class="mi">2</span><span class="p">)]</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-19'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-19'>#</a>
|
||||
</div>
|
||||
<p>使用层创建顺序模型</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">103</span> <span class="bp">self</span><span class="o">.</span><span class="n">layers</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Sequential</span><span class="p">(</span><span class="o">*</span><span class="n">layers</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-20'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-20'>#</a>
|
||||
</div>
|
||||
<p>最后的 logits 层</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">105</span> <span class="bp">self</span><span class="o">.</span><span class="n">fc</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Linear</span><span class="p">(</span><span class="n">in_channels</span><span class="p">,</span> <span class="mi">10</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-21'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-21'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">107</span> <span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-22'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-22'>#</a>
|
||||
</div>
|
||||
<p>VGG 层</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">109</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">layers</span><span class="p">(</span><span class="n">x</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-23'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-23'>#</a>
|
||||
</div>
|
||||
<p>修改分类图层的形状</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">111</span> <span class="n">x</span> <span class="o">=</span> <span class="n">x</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="n">x</span><span class="o">.</span><span class="n">shape</span><span class="p">[</span><span class="mi">0</span><span class="p">],</span> <span class="o">-</span><span class="mi">1</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-24'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-24'>#</a>
|
||||
</div>
|
||||
<p>最后的线性层</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">113</span> <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">fc</span><span class="p">(</span><span class="n">x</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
@@ -0,0 +1,113 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content=""/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="None"/>
|
||||
<meta name="twitter:description" content=""/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/experiments/index.html"/>
|
||||
<meta property="og:title" content="None"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="None"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="None"/>
|
||||
<meta property="og:description" content=""/>
|
||||
|
||||
<title>None</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/experiments/index.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="index.html">experiments</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/experiments/__init__.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
@@ -0,0 +1,480 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content="这是 MNIST 数据集的可重复使用的训练器"/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="MNIST 实验"/>
|
||||
<meta name="twitter:description" content="这是 MNIST 数据集的可重复使用的训练器"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/experiments/mnist.html"/>
|
||||
<meta property="og:title" content="MNIST 实验"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="MNIST 实验"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="MNIST 实验"/>
|
||||
<meta property="og:description" content="这是 MNIST 数据集的可重复使用的训练器"/>
|
||||
|
||||
<title>MNIST 实验</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/experiments/mnist.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="index.html">experiments</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/experiments/mnist.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1>MNIST 实验</h1>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">11</span><span></span><span class="kn">import</span> <span class="nn">torch.nn</span> <span class="k">as</span> <span class="nn">nn</span>
|
||||
<span class="lineno">12</span><span class="kn">import</span> <span class="nn">torch.utils.data</span>
|
||||
<span class="lineno">13</span><span class="kn">from</span> <span class="nn">labml_helpers.module</span> <span class="kn">import</span> <span class="n">Module</span>
|
||||
<span class="lineno">14</span>
|
||||
<span class="lineno">15</span><span class="kn">from</span> <span class="nn">labml</span> <span class="kn">import</span> <span class="n">tracker</span>
|
||||
<span class="lineno">16</span><span class="kn">from</span> <span class="nn">labml.configs</span> <span class="kn">import</span> <span class="n">option</span>
|
||||
<span class="lineno">17</span><span class="kn">from</span> <span class="nn">labml_helpers.datasets.mnist</span> <span class="kn">import</span> <span class="n">MNISTConfigs</span> <span class="k">as</span> <span class="n">MNISTDatasetConfigs</span>
|
||||
<span class="lineno">18</span><span class="kn">from</span> <span class="nn">labml_helpers.device</span> <span class="kn">import</span> <span class="n">DeviceConfigs</span>
|
||||
<span class="lineno">19</span><span class="kn">from</span> <span class="nn">labml_helpers.metrics.accuracy</span> <span class="kn">import</span> <span class="n">Accuracy</span>
|
||||
<span class="lineno">20</span><span class="kn">from</span> <span class="nn">labml_helpers.train_valid</span> <span class="kn">import</span> <span class="n">TrainValidConfigs</span><span class="p">,</span> <span class="n">BatchIndex</span><span class="p">,</span> <span class="n">hook_model_outputs</span>
|
||||
<span class="lineno">21</span><span class="kn">from</span> <span class="nn">labml_nn.optimizers.configs</span> <span class="kn">import</span> <span class="n">OptimizerConfigs</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-1'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-1'>#</a>
|
||||
</div>
|
||||
<p><a id="MNISTConfigs"></a></p>
|
||||
<h2>训练器配置</h2>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">24</span><span class="k">class</span> <span class="nc">MNISTConfigs</span><span class="p">(</span><span class="n">MNISTDatasetConfigs</span><span class="p">,</span> <span class="n">TrainValidConfigs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-2'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-2'>#</a>
|
||||
</div>
|
||||
<p>优化器</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">32</span> <span class="n">optimizer</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">optim</span><span class="o">.</span><span class="n">Adam</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-3'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-3'>#</a>
|
||||
</div>
|
||||
<p>训练设备</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">34</span> <span class="n">device</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">device</span> <span class="o">=</span> <span class="n">DeviceConfigs</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-4'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-4'>#</a>
|
||||
</div>
|
||||
<p>分类模型</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">37</span> <span class="n">model</span><span class="p">:</span> <span class="n">Module</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-5'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-5'>#</a>
|
||||
</div>
|
||||
<p>要训练的时代数</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">39</span> <span class="n">epochs</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">10</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-6'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-6'>#</a>
|
||||
</div>
|
||||
<p>一个纪元内在训练和验证之间切换的次数</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">42</span> <span class="n">inner_iterations</span> <span class="o">=</span> <span class="mi">10</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-7'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-7'>#</a>
|
||||
</div>
|
||||
<p>精度函数</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">45</span> <span class="n">accuracy</span> <span class="o">=</span> <span class="n">Accuracy</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-8'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-8'>#</a>
|
||||
</div>
|
||||
<p>亏损函数</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">47</span> <span class="n">loss_func</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">CrossEntropyLoss</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-9'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-9'>#</a>
|
||||
</div>
|
||||
<h3>初始化</h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">49</span> <span class="k">def</span> <span class="nf">init</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-10'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-10'>#</a>
|
||||
</div>
|
||||
<p>设置跟踪器配置</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">54</span> <span class="n">tracker</span><span class="o">.</span><span class="n">set_scalar</span><span class="p">(</span><span class="s2">"loss.*"</span><span class="p">,</span> <span class="kc">True</span><span class="p">)</span>
|
||||
<span class="lineno">55</span> <span class="n">tracker</span><span class="o">.</span><span class="n">set_scalar</span><span class="p">(</span><span class="s2">"accuracy.*"</span><span class="p">,</span> <span class="kc">True</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-11'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-11'>#</a>
|
||||
</div>
|
||||
<p>向日志模块输出添加钩子</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">57</span> <span class="n">hook_model_outputs</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">mode</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">model</span><span class="p">,</span> <span class="s1">'model'</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-12'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-12'>#</a>
|
||||
</div>
|
||||
<p>增加作为状态模块的精度。这个名字可能令人困惑，因为它旨在存储 RNN 的训练和验证之间的状态。这将使精度指标统计数据分开，以便进行训练和验证。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">62</span> <span class="bp">self</span><span class="o">.</span><span class="n">state_modules</span> <span class="o">=</span> <span class="p">[</span><span class="bp">self</span><span class="o">.</span><span class="n">accuracy</span><span class="p">]</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-13'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-13'>#</a>
|
||||
</div>
|
||||
<h3>培训或验证步骤</h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">64</span> <span class="k">def</span> <span class="nf">step</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">batch</span><span class="p">:</span> <span class="nb">any</span><span class="p">,</span> <span class="n">batch_idx</span><span class="p">:</span> <span class="n">BatchIndex</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-14'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-14'>#</a>
|
||||
</div>
|
||||
<p>训练/评估模式</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">70</span> <span class="bp">self</span><span class="o">.</span><span class="n">model</span><span class="o">.</span><span class="n">train</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">mode</span><span class="o">.</span><span class="n">is_train</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-15'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-15'>#</a>
|
||||
</div>
|
||||
<p>将数据移动到设备</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">73</span> <span class="n">data</span><span class="p">,</span> <span class="n">target</span> <span class="o">=</span> <span class="n">batch</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">device</span><span class="p">),</span> <span class="n">batch</span><span class="p">[</span><span class="mi">1</span><span class="p">]</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">device</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-16'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-16'>#</a>
|
||||
</div>
|
||||
<p>在训练模式下更新全局步长（处理的样本数）</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">76</span> <span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">mode</span><span class="o">.</span><span class="n">is_train</span><span class="p">:</span>
|
||||
<span class="lineno">77</span> <span class="n">tracker</span><span class="o">.</span><span class="n">add_global_step</span><span class="p">(</span><span class="nb">len</span><span class="p">(</span><span class="n">data</span><span class="p">))</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-17'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-17'>#</a>
|
||||
</div>
|
||||
<p>是否捕获模型输出</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">80</span> <span class="k">with</span> <span class="bp">self</span><span class="o">.</span><span class="n">mode</span><span class="o">.</span><span class="n">update</span><span class="p">(</span><span class="n">is_log_activations</span><span class="o">=</span><span class="n">batch_idx</span><span class="o">.</span><span class="n">is_last</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-18'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-18'>#</a>
|
||||
</div>
|
||||
<p>获取模型输出。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">82</span> <span class="n">output</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">model</span><span class="p">(</span><span class="n">data</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-19'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-19'>#</a>
|
||||
</div>
|
||||
<p>计算并记录损失</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">85</span> <span class="n">loss</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">loss_func</span><span class="p">(</span><span class="n">output</span><span class="p">,</span> <span class="n">target</span><span class="p">)</span>
|
||||
<span class="lineno">86</span> <span class="n">tracker</span><span class="o">.</span><span class="n">add</span><span class="p">(</span><span class="s2">"loss."</span><span class="p">,</span> <span class="n">loss</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-20'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-20'>#</a>
|
||||
</div>
|
||||
<p>计算和记录精度</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">89</span> <span class="bp">self</span><span class="o">.</span><span class="n">accuracy</span><span class="p">(</span><span class="n">output</span><span class="p">,</span> <span class="n">target</span><span class="p">)</span>
|
||||
<span class="lineno">90</span> <span class="bp">self</span><span class="o">.</span><span class="n">accuracy</span><span class="o">.</span><span class="n">track</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-21'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-21'>#</a>
|
||||
</div>
|
||||
<p>训练模型</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">93</span> <span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">mode</span><span class="o">.</span><span class="n">is_train</span><span class="p">:</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-22'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-22'>#</a>
|
||||
</div>
|
||||
<p>计算梯度</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">95</span> <span class="n">loss</span><span class="o">.</span><span class="n">backward</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-23'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-23'>#</a>
|
||||
</div>
|
||||
<p>采取优化器步骤</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">97</span> <span class="bp">self</span><span class="o">.</span><span class="n">optimizer</span><span class="o">.</span><span class="n">step</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-24'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-24'>#</a>
|
||||
</div>
|
||||
<p>记录每个纪元最后一批的模型参数和梯度</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">99</span> <span class="k">if</span> <span class="n">batch_idx</span><span class="o">.</span><span class="n">is_last</span><span class="p">:</span>
|
||||
<span class="lineno">100</span> <span class="n">tracker</span><span class="o">.</span><span class="n">add</span><span class="p">(</span><span class="s1">'model'</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">model</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-25'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-25'>#</a>
|
||||
</div>
|
||||
<p>清除渐变</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">102</span> <span class="bp">self</span><span class="o">.</span><span class="n">optimizer</span><span class="o">.</span><span class="n">zero_grad</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-26'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-26'>#</a>
|
||||
</div>
|
||||
<p>保存跟踪的指标</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">105</span> <span class="n">tracker</span><span class="o">.</span><span class="n">save</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-27'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-27'>#</a>
|
||||
</div>
|
||||
<h3>默认优化器配置</h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">108</span><span class="nd">@option</span><span class="p">(</span><span class="n">MNISTConfigs</span><span class="o">.</span><span class="n">optimizer</span><span class="p">)</span>
|
||||
<span class="lineno">109</span><span class="k">def</span> <span class="nf">_optimizer</span><span class="p">(</span><span class="n">c</span><span class="p">:</span> <span class="n">MNISTConfigs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-28'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-28'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">113</span> <span class="n">opt_conf</span> <span class="o">=</span> <span class="n">OptimizerConfigs</span><span class="p">()</span>
|
||||
<span class="lineno">114</span> <span class="n">opt_conf</span><span class="o">.</span><span class="n">parameters</span> <span class="o">=</span> <span class="n">c</span><span class="o">.</span><span class="n">model</span><span class="o">.</span><span class="n">parameters</span><span class="p">()</span>
|
||||
<span class="lineno">115</span> <span class="n">opt_conf</span><span class="o">.</span><span class="n">optimizer</span> <span class="o">=</span> <span class="s1">'Adam'</span>
|
||||
<span class="lineno">116</span> <span class="k">return</span> <span class="n">opt_conf</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
File diff suppressed because it is too large
Load Diff
File diff suppressed because it is too large
Load Diff
File diff suppressed because one or more lines are too long
@@ -0,0 +1,127 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content=""/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="循环增益"/>
|
||||
<meta name="twitter:description" content=""/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/gan/cycle_gan/readme.html"/>
|
||||
<meta property="og:title" content="循环增益"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="循环增益"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="循环增益"/>
|
||||
<meta property="og:description" content=""/>
|
||||
|
||||
<title>循环增益</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/gan/cycle_gan/readme.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="../index.html">gan</a>
|
||||
<a class="parent" href="index.html">cycle_gan</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/gan/cycle_gan/readme.md" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1><a href="https://nn.labml.ai/gan/cycle_gan/index.html">循环增益</a></h1>
|
||||
<p>这是论文《<a href="https://arxiv.org/abs/1703.10593">使用周期一致的对抗网络进行未配对的图像到图像转换》的 Py</a> <a href="https://pytorch.org">Torch</a> 实现/教程。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
@@ -0,0 +1,413 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content="深度卷积生成对抗网络（DCGAN）的简单的 PyTorch 实现/教程。"/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="深度卷积生成对抗网络 (DCGAN)"/>
|
||||
<meta name="twitter:description" content="深度卷积生成对抗网络（DCGAN）的简单的 PyTorch 实现/教程。"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/gan/dcgan/index.html"/>
|
||||
<meta property="og:title" content="深度卷积生成对抗网络 (DCGAN)"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="深度卷积生成对抗网络 (DCGAN)"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="深度卷积生成对抗网络 (DCGAN)"/>
|
||||
<meta property="og:description" content="深度卷积生成对抗网络（DCGAN）的简单的 PyTorch 实现/教程。"/>
|
||||
|
||||
<title>深度卷积生成对抗网络 (DCGAN)</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/gan/dcgan/index.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="../index.html">gan</a>
|
||||
<a class="parent" href="index.html">dcgan</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/gan/dcgan/__init__.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1>深度卷积生成对抗网络 (DCGAN)</h1>
|
||||
<p>这是 <a href="https://pytorch.org">PyTorch</a> 实现的论文《<a href="https://arxiv.org/abs/1511.06434">使用深度卷积生成对抗网络进行无监督表示学习</a>》。</p>
|
||||
<p>此实现基于 <a href="https://pytorch.org/tutorials/beginner/dcgan_faces_tutorial.html">PyTorch DCGAN 教程</a>。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">15</span><span></span><span class="kn">import</span> <span class="nn">torch.nn</span> <span class="k">as</span> <span class="nn">nn</span>
|
||||
<span class="lineno">16</span>
|
||||
<span class="lineno">17</span><span class="kn">from</span> <span class="nn">labml</span> <span class="kn">import</span> <span class="n">experiment</span>
|
||||
<span class="lineno">18</span><span class="kn">from</span> <span class="nn">labml.configs</span> <span class="kn">import</span> <span class="n">calculate</span>
|
||||
<span class="lineno">19</span><span class="kn">from</span> <span class="nn">labml_helpers.module</span> <span class="kn">import</span> <span class="n">Module</span>
|
||||
<span class="lineno">20</span><span class="kn">from</span> <span class="nn">labml_nn.gan.original.experiment</span> <span class="kn">import</span> <span class="n">Configs</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-1'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-1'>#</a>
|
||||
</div>
|
||||
<h3>卷积生成器网络</h3>
|
||||
<p>这类似于用于 CeleBA 人脸的反卷积网络，但针对 MNIST 图像进行了修改。</p>
|
||||
<p><img alt="DCGan Architecture" src="https://pytorch.org/tutorials/_images/dcgan_generator.png"></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">23</span><span class="k">class</span> <span class="nc">Generator</span><span class="p">(</span><span class="n">Module</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-2'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-2'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">33</span> <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
|
||||
<span class="lineno">34</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-3'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-3'>#</a>
|
||||
</div>
|
||||
<p>输入<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.72777em;vertical-align:-0.08333em;"></span><span class="mord coloredeq eqc" style=""><span class="mord" style="">1</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">×</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord" style="">1</span></span></span></span></span></span>有 100 个通道</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">36</span> <span class="bp">self</span><span class="o">.</span><span class="n">layers</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Sequential</span><span class="p">(</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-4'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-4'>#</a>
|
||||
</div>
|
||||
<p>这给出了<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.72777em;vertical-align:-0.08333em;"></span><span class="mord coloredeq eqd" style=""><span class="mord" style="">3</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">×</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord" style="">3</span></span></span></span></span></span>输出</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">38</span> <span class="n">nn</span><span class="o">.</span><span class="n">ConvTranspose2d</span><span class="p">(</span><span class="mi">100</span><span class="p">,</span> <span class="mi">1024</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">0</span><span class="p">,</span> <span class="n">bias</span><span class="o">=</span><span class="kc">False</span><span class="p">),</span>
|
||||
<span class="lineno">39</span> <span class="n">nn</span><span class="o">.</span><span class="n">BatchNorm2d</span><span class="p">(</span><span class="mi">1024</span><span class="p">),</span>
|
||||
<span class="lineno">40</span> <span class="n">nn</span><span class="o">.</span><span class="n">ReLU</span><span class="p">(</span><span class="kc">True</span><span class="p">),</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-5'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-5'>#</a>
|
||||
</div>
|
||||
<p>这给了<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.72777em;vertical-align:-0.08333em;"></span><span class="mord coloredeq eqe" style=""><span class="mord" style="">7</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">×</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord" style="">7</span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">42</span> <span class="n">nn</span><span class="o">.</span><span class="n">ConvTranspose2d</span><span class="p">(</span><span class="mi">1024</span><span class="p">,</span> <span class="mi">512</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="mi">0</span><span class="p">,</span> <span class="n">bias</span><span class="o">=</span><span class="kc">False</span><span class="p">),</span>
|
||||
<span class="lineno">43</span> <span class="n">nn</span><span class="o">.</span><span class="n">BatchNorm2d</span><span class="p">(</span><span class="mi">512</span><span class="p">),</span>
|
||||
<span class="lineno">44</span> <span class="n">nn</span><span class="o">.</span><span class="n">ReLU</span><span class="p">(</span><span class="kc">True</span><span class="p">),</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-6'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-6'>#</a>
|
||||
</div>
|
||||
<p>这给了<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.72777em;vertical-align:-0.08333em;"></span><span class="mord coloredeq eqa" style=""><span class="mord" style="">14</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">×</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord" style="">14</span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">46</span> <span class="n">nn</span><span class="o">.</span><span class="n">ConvTranspose2d</span><span class="p">(</span><span class="mi">512</span><span class="p">,</span> <span class="mi">256</span><span class="p">,</span> <span class="mi">4</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="n">bias</span><span class="o">=</span><span class="kc">False</span><span class="p">),</span>
|
||||
<span class="lineno">47</span> <span class="n">nn</span><span class="o">.</span><span class="n">BatchNorm2d</span><span class="p">(</span><span class="mi">256</span><span class="p">),</span>
|
||||
<span class="lineno">48</span> <span class="n">nn</span><span class="o">.</span><span class="n">ReLU</span><span class="p">(</span><span class="kc">True</span><span class="p">),</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-7'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-7'>#</a>
|
||||
</div>
|
||||
<p>这给了<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.72777em;vertical-align:-0.08333em;"></span><span class="mord coloredeq eqb" style=""><span class="mord" style="">28</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">×</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord" style="">28</span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">50</span> <span class="n">nn</span><span class="o">.</span><span class="n">ConvTranspose2d</span><span class="p">(</span><span class="mi">256</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">4</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="n">bias</span><span class="o">=</span><span class="kc">False</span><span class="p">),</span>
|
||||
<span class="lineno">51</span> <span class="n">nn</span><span class="o">.</span><span class="n">Tanh</span><span class="p">()</span>
|
||||
<span class="lineno">52</span> <span class="p">)</span>
|
||||
<span class="lineno">53</span>
|
||||
<span class="lineno">54</span> <span class="bp">self</span><span class="o">.</span><span class="n">apply</span><span class="p">(</span><span class="n">_weights_init</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-8'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-8'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">56</span> <span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-9'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-9'>#</a>
|
||||
</div>
|
||||
<p>从形状改<code class="highlight"><span></span><span class="p">[</span><span class="n">batch_size</span><span class="p">,</span> <span class="mi">100</span><span class="p">]</span></code>
|
||||
为<code class="highlight"><span></span><span class="p">[</span><span class="n">batch_size</span><span class="p">,</span> <span class="mi">100</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">]</span></code>
|
||||
</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">58</span> <span class="n">x</span> <span class="o">=</span> <span class="n">x</span><span class="o">.</span><span class="n">unsqueeze</span><span class="p">(</span><span class="o">-</span><span class="mi">1</span><span class="p">)</span><span class="o">.</span><span class="n">unsqueeze</span><span class="p">(</span><span class="o">-</span><span class="mi">1</span><span class="p">)</span>
|
||||
<span class="lineno">59</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">layers</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
|
||||
<span class="lineno">60</span> <span class="k">return</span> <span class="n">x</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-10'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-10'>#</a>
|
||||
</div>
|
||||
<h3>卷积鉴别器网络</h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">63</span><span class="k">class</span> <span class="nc">Discriminator</span><span class="p">(</span><span class="n">Module</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-11'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-11'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">68</span> <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
|
||||
<span class="lineno">69</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-12'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-12'>#</a>
|
||||
</div>
|
||||
<p>输入<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.72777em;vertical-align:-0.08333em;"></span><span class="mord coloredeq eqb" style=""><span class="mord" style="">28</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">×</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord" style="">28</span></span></span></span></span></span>使用一个通道</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">71</span> <span class="bp">self</span><span class="o">.</span><span class="n">layers</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Sequential</span><span class="p">(</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-13'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-13'>#</a>
|
||||
</div>
|
||||
<p>这给了<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.72777em;vertical-align:-0.08333em;"></span><span class="mord coloredeq eqa" style=""><span class="mord" style="">14</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">×</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord" style="">14</span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">73</span> <span class="n">nn</span><span class="o">.</span><span class="n">Conv2d</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="mi">256</span><span class="p">,</span> <span class="mi">4</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="n">bias</span><span class="o">=</span><span class="kc">False</span><span class="p">),</span>
|
||||
<span class="lineno">74</span> <span class="n">nn</span><span class="o">.</span><span class="n">LeakyReLU</span><span class="p">(</span><span class="mf">0.2</span><span class="p">,</span> <span class="n">inplace</span><span class="o">=</span><span class="kc">True</span><span class="p">),</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-14'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-14'>#</a>
|
||||
</div>
|
||||
<p>这给了<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.72777em;vertical-align:-0.08333em;"></span><span class="mord coloredeq eqe" style=""><span class="mord" style="">7</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">×</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord" style="">7</span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">76</span> <span class="n">nn</span><span class="o">.</span><span class="n">Conv2d</span><span class="p">(</span><span class="mi">256</span><span class="p">,</span> <span class="mi">512</span><span class="p">,</span> <span class="mi">4</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="n">bias</span><span class="o">=</span><span class="kc">False</span><span class="p">),</span>
|
||||
<span class="lineno">77</span> <span class="n">nn</span><span class="o">.</span><span class="n">BatchNorm2d</span><span class="p">(</span><span class="mi">512</span><span class="p">),</span>
|
||||
<span class="lineno">78</span> <span class="n">nn</span><span class="o">.</span><span class="n">LeakyReLU</span><span class="p">(</span><span class="mf">0.2</span><span class="p">,</span> <span class="n">inplace</span><span class="o">=</span><span class="kc">True</span><span class="p">),</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-15'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-15'>#</a>
|
||||
</div>
|
||||
<p>这给了<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.72777em;vertical-align:-0.08333em;"></span><span class="mord coloredeq eqd" style=""><span class="mord" style="">3</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">×</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord" style="">3</span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">80</span> <span class="n">nn</span><span class="o">.</span><span class="n">Conv2d</span><span class="p">(</span><span class="mi">512</span><span class="p">,</span> <span class="mi">1024</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="mi">0</span><span class="p">,</span> <span class="n">bias</span><span class="o">=</span><span class="kc">False</span><span class="p">),</span>
|
||||
<span class="lineno">81</span> <span class="n">nn</span><span class="o">.</span><span class="n">BatchNorm2d</span><span class="p">(</span><span class="mi">1024</span><span class="p">),</span>
|
||||
<span class="lineno">82</span> <span class="n">nn</span><span class="o">.</span><span class="n">LeakyReLU</span><span class="p">(</span><span class="mf">0.2</span><span class="p">,</span> <span class="n">inplace</span><span class="o">=</span><span class="kc">True</span><span class="p">),</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-16'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-16'>#</a>
|
||||
</div>
|
||||
<p>这给了<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.72777em;vertical-align:-0.08333em;"></span><span class="mord coloredeq eqc" style=""><span class="mord" style="">1</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mbin" style="">×</span><span class="mspace" style="margin-right:0.2222222222222222em;"></span><span class="mord" style="">1</span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">84</span> <span class="n">nn</span><span class="o">.</span><span class="n">Conv2d</span><span class="p">(</span><span class="mi">1024</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">0</span><span class="p">,</span> <span class="n">bias</span><span class="o">=</span><span class="kc">False</span><span class="p">),</span>
|
||||
<span class="lineno">85</span> <span class="p">)</span>
|
||||
<span class="lineno">86</span> <span class="bp">self</span><span class="o">.</span><span class="n">apply</span><span class="p">(</span><span class="n">_weights_init</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-17'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-17'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">88</span> <span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">):</span>
|
||||
<span class="lineno">89</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">layers</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
|
||||
<span class="lineno">90</span> <span class="k">return</span> <span class="n">x</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="n">x</span><span class="o">.</span><span class="n">shape</span><span class="p">[</span><span class="mi">0</span><span class="p">],</span> <span class="o">-</span><span class="mi">1</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-18'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-18'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">93</span><span class="k">def</span> <span class="nf">_weights_init</span><span class="p">(</span><span class="n">m</span><span class="p">):</span>
|
||||
<span class="lineno">94</span> <span class="n">classname</span> <span class="o">=</span> <span class="n">m</span><span class="o">.</span><span class="vm">__class__</span><span class="o">.</span><span class="vm">__name__</span>
|
||||
<span class="lineno">95</span> <span class="k">if</span> <span class="n">classname</span><span class="o">.</span><span class="n">find</span><span class="p">(</span><span class="s1">'Conv'</span><span class="p">)</span> <span class="o">!=</span> <span class="o">-</span><span class="mi">1</span><span class="p">:</span>
|
||||
<span class="lineno">96</span> <span class="n">nn</span><span class="o">.</span><span class="n">init</span><span class="o">.</span><span class="n">normal_</span><span class="p">(</span><span class="n">m</span><span class="o">.</span><span class="n">weight</span><span class="o">.</span><span class="n">data</span><span class="p">,</span> <span class="mf">0.0</span><span class="p">,</span> <span class="mf">0.02</span><span class="p">)</span>
|
||||
<span class="lineno">97</span> <span class="k">elif</span> <span class="n">classname</span><span class="o">.</span><span class="n">find</span><span class="p">(</span><span class="s1">'BatchNorm'</span><span class="p">)</span> <span class="o">!=</span> <span class="o">-</span><span class="mi">1</span><span class="p">:</span>
|
||||
<span class="lineno">98</span> <span class="n">nn</span><span class="o">.</span><span class="n">init</span><span class="o">.</span><span class="n">normal_</span><span class="p">(</span><span class="n">m</span><span class="o">.</span><span class="n">weight</span><span class="o">.</span><span class="n">data</span><span class="p">,</span> <span class="mf">1.0</span><span class="p">,</span> <span class="mf">0.02</span><span class="p">)</span>
|
||||
<span class="lineno">99</span> <span class="n">nn</span><span class="o">.</span><span class="n">init</span><span class="o">.</span><span class="n">constant_</span><span class="p">(</span><span class="n">m</span><span class="o">.</span><span class="n">bias</span><span class="o">.</span><span class="n">data</span><span class="p">,</span> <span class="mi">0</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-19'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-19'>#</a>
|
||||
</div>
|
||||
<p>我们导入了<a href="../original/experiment.html">简单的 gan 实验</a>并更改了生成器和鉴别器网络</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">104</span><span class="n">calculate</span><span class="p">(</span><span class="n">Configs</span><span class="o">.</span><span class="n">generator</span><span class="p">,</span> <span class="s1">'cnn'</span><span class="p">,</span> <span class="k">lambda</span> <span class="n">c</span><span class="p">:</span> <span class="n">Generator</span><span class="p">()</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">device</span><span class="p">))</span>
|
||||
<span class="lineno">105</span><span class="n">calculate</span><span class="p">(</span><span class="n">Configs</span><span class="o">.</span><span class="n">discriminator</span><span class="p">,</span> <span class="s1">'cnn'</span><span class="p">,</span> <span class="k">lambda</span> <span class="n">c</span><span class="p">:</span> <span class="n">Discriminator</span><span class="p">()</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">device</span><span class="p">))</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-20'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-20'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">108</span><span class="k">def</span> <span class="nf">main</span><span class="p">():</span>
|
||||
<span class="lineno">109</span> <span class="n">conf</span> <span class="o">=</span> <span class="n">Configs</span><span class="p">()</span>
|
||||
<span class="lineno">110</span> <span class="n">experiment</span><span class="o">.</span><span class="n">create</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">'mnist_dcgan'</span><span class="p">)</span>
|
||||
<span class="lineno">111</span> <span class="n">experiment</span><span class="o">.</span><span class="n">configs</span><span class="p">(</span><span class="n">conf</span><span class="p">,</span>
|
||||
<span class="lineno">112</span> <span class="p">{</span><span class="s1">'discriminator'</span><span class="p">:</span> <span class="s1">'cnn'</span><span class="p">,</span>
|
||||
<span class="lineno">113</span> <span class="s1">'generator'</span><span class="p">:</span> <span class="s1">'cnn'</span><span class="p">,</span>
|
||||
<span class="lineno">114</span> <span class="s1">'label_smoothing'</span><span class="p">:</span> <span class="mf">0.01</span><span class="p">})</span>
|
||||
<span class="lineno">115</span> <span class="k">with</span> <span class="n">experiment</span><span class="o">.</span><span class="n">start</span><span class="p">():</span>
|
||||
<span class="lineno">116</span> <span class="n">conf</span><span class="o">.</span><span class="n">run</span><span class="p">()</span>
|
||||
<span class="lineno">117</span>
|
||||
<span class="lineno">118</span>
|
||||
<span class="lineno">119</span><span class="k">if</span> <span class="vm">__name__</span> <span class="o">==</span> <span class="s1">'__main__'</span><span class="p">:</span>
|
||||
<span class="lineno">120</span> <span class="n">main</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
@@ -0,0 +1,127 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content=""/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="深度卷积生成对抗网络-DCGAN"/>
|
||||
<meta name="twitter:description" content=""/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/gan/dcgan/readme.html"/>
|
||||
<meta property="og:title" content="深度卷积生成对抗网络-DCGAN"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="深度卷积生成对抗网络-DCGAN"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="深度卷积生成对抗网络-DCGAN"/>
|
||||
<meta property="og:description" content=""/>
|
||||
|
||||
<title>深度卷积生成对抗网络-DCGAN</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/gan/dcgan/readme.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="../index.html">gan</a>
|
||||
<a class="parent" href="index.html">dcgan</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/gan/dcgan/readme.md" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1><a href="https://nn.labml.ai/gan/dcgan/index.html">深度卷积生成对抗网络-DCGAN</a></h1>
|
||||
<p>这是 <a href="https://pytorch.org">PyTorch</a> 实现的论文《<a href="https://arxiv.org/abs/1511.06434">使用深度卷积生成对抗网络进行无监督表示学习</a>》。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
@@ -0,0 +1,131 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content="一组 GaN 的 PyTorch 实现/教程。"/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="生成对抗网络"/>
|
||||
<meta name="twitter:description" content="一组 GaN 的 PyTorch 实现/教程。"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/gan/index.html"/>
|
||||
<meta property="og:title" content="生成对抗网络"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="生成对抗网络"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="生成对抗网络"/>
|
||||
<meta property="og:description" content="一组 GaN 的 PyTorch 实现/教程。"/>
|
||||
|
||||
<title>生成对抗网络</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/gan/index.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="index.html">gan</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/gan/__init__.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1>生成对抗网络</h1>
|
||||
<ul><li><a href="original/index.html">原装 GAN</a></li>
|
||||
<li><a href="dcgan/index.html">具有深度卷积网络的 GAN</a></li>
|
||||
<li><a href="cycle_gan/index.html">循环增益</a></li>
|
||||
<li><a href="wasserstein/index.html">Wasserstein GAN</a></li>
|
||||
<li><a href="wasserstein/gradient_penalty/index.html">Wasserstein GAN 带梯度惩罚</a></li>
|
||||
<li><a href="stylegan/index.html">StyleGan 2</a></li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
@@ -0,0 +1,646 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content="该实验使用多层感知器生成 MNIST 图像。"/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="使用 MNIST 进行生成对抗网络实验"/>
|
||||
<meta name="twitter:description" content="该实验使用多层感知器生成 MNIST 图像。"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/gan/original/experiment.html"/>
|
||||
<meta property="og:title" content="使用 MNIST 进行生成对抗网络实验"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="使用 MNIST 进行生成对抗网络实验"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="使用 MNIST 进行生成对抗网络实验"/>
|
||||
<meta property="og:description" content="该实验使用多层感知器生成 MNIST 图像。"/>
|
||||
|
||||
<title>使用 MNIST 进行生成对抗网络实验</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/gan/original/experiment.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="../index.html">gan</a>
|
||||
<a class="parent" href="index.html">original</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/gan/original/experiment.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1>使用 MNIST 进行生成对抗网络实验</h1>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">10</span><span></span><span class="kn">from</span> <span class="nn">typing</span> <span class="kn">import</span> <span class="n">Any</span>
|
||||
<span class="lineno">11</span>
|
||||
<span class="lineno">12</span><span class="kn">import</span> <span class="nn">torch</span>
|
||||
<span class="lineno">13</span><span class="kn">import</span> <span class="nn">torch.nn</span> <span class="k">as</span> <span class="nn">nn</span>
|
||||
<span class="lineno">14</span><span class="kn">import</span> <span class="nn">torch.utils.data</span>
|
||||
<span class="lineno">15</span><span class="kn">from</span> <span class="nn">torchvision</span> <span class="kn">import</span> <span class="n">transforms</span>
|
||||
<span class="lineno">16</span>
|
||||
<span class="lineno">17</span><span class="kn">from</span> <span class="nn">labml</span> <span class="kn">import</span> <span class="n">tracker</span><span class="p">,</span> <span class="n">monit</span><span class="p">,</span> <span class="n">experiment</span>
|
||||
<span class="lineno">18</span><span class="kn">from</span> <span class="nn">labml.configs</span> <span class="kn">import</span> <span class="n">option</span><span class="p">,</span> <span class="n">calculate</span>
|
||||
<span class="lineno">19</span><span class="kn">from</span> <span class="nn">labml_helpers.datasets.mnist</span> <span class="kn">import</span> <span class="n">MNISTConfigs</span>
|
||||
<span class="lineno">20</span><span class="kn">from</span> <span class="nn">labml_helpers.device</span> <span class="kn">import</span> <span class="n">DeviceConfigs</span>
|
||||
<span class="lineno">21</span><span class="kn">from</span> <span class="nn">labml_helpers.module</span> <span class="kn">import</span> <span class="n">Module</span>
|
||||
<span class="lineno">22</span><span class="kn">from</span> <span class="nn">labml_helpers.optimizer</span> <span class="kn">import</span> <span class="n">OptimizerConfigs</span>
|
||||
<span class="lineno">23</span><span class="kn">from</span> <span class="nn">labml_helpers.train_valid</span> <span class="kn">import</span> <span class="n">TrainValidConfigs</span><span class="p">,</span> <span class="n">hook_model_outputs</span><span class="p">,</span> <span class="n">BatchIndex</span>
|
||||
<span class="lineno">24</span><span class="kn">from</span> <span class="nn">labml_nn.gan.original</span> <span class="kn">import</span> <span class="n">DiscriminatorLogitsLoss</span><span class="p">,</span> <span class="n">GeneratorLogitsLoss</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-1'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-1'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">27</span><span class="k">def</span> <span class="nf">weights_init</span><span class="p">(</span><span class="n">m</span><span class="p">):</span>
|
||||
<span class="lineno">28</span> <span class="n">classname</span> <span class="o">=</span> <span class="n">m</span><span class="o">.</span><span class="vm">__class__</span><span class="o">.</span><span class="vm">__name__</span>
|
||||
<span class="lineno">29</span> <span class="k">if</span> <span class="n">classname</span><span class="o">.</span><span class="n">find</span><span class="p">(</span><span class="s1">'Linear'</span><span class="p">)</span> <span class="o">!=</span> <span class="o">-</span><span class="mi">1</span><span class="p">:</span>
|
||||
<span class="lineno">30</span> <span class="n">nn</span><span class="o">.</span><span class="n">init</span><span class="o">.</span><span class="n">normal_</span><span class="p">(</span><span class="n">m</span><span class="o">.</span><span class="n">weight</span><span class="o">.</span><span class="n">data</span><span class="p">,</span> <span class="mf">0.0</span><span class="p">,</span> <span class="mf">0.02</span><span class="p">)</span>
|
||||
<span class="lineno">31</span> <span class="k">elif</span> <span class="n">classname</span><span class="o">.</span><span class="n">find</span><span class="p">(</span><span class="s1">'BatchNorm'</span><span class="p">)</span> <span class="o">!=</span> <span class="o">-</span><span class="mi">1</span><span class="p">:</span>
|
||||
<span class="lineno">32</span> <span class="n">nn</span><span class="o">.</span><span class="n">init</span><span class="o">.</span><span class="n">normal_</span><span class="p">(</span><span class="n">m</span><span class="o">.</span><span class="n">weight</span><span class="o">.</span><span class="n">data</span><span class="p">,</span> <span class="mf">1.0</span><span class="p">,</span> <span class="mf">0.02</span><span class="p">)</span>
|
||||
<span class="lineno">33</span> <span class="n">nn</span><span class="o">.</span><span class="n">init</span><span class="o">.</span><span class="n">constant_</span><span class="p">(</span><span class="n">m</span><span class="o">.</span><span class="n">bias</span><span class="o">.</span><span class="n">data</span><span class="p">,</span> <span class="mi">0</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-2'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-2'>#</a>
|
||||
</div>
|
||||
<h3>简单的 MLP 生成器</h3>
|
||||
<p>它有三个线性层，随着<code class="highlight"><span></span><span class="n">LeakyReLU</span></code>
|
||||
激活的大小不断增加。最后一层已激<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.69444em;vertical-align:0em;"></span><span class="mord mathnormal">t</span><span class="mord mathnormal">anh</span></span></span></span></span>活。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">36</span><span class="k">class</span> <span class="nc">Generator</span><span class="p">(</span><span class="n">Module</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-3'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-3'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">44</span> <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
|
||||
<span class="lineno">45</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
|
||||
<span class="lineno">46</span> <span class="n">layer_sizes</span> <span class="o">=</span> <span class="p">[</span><span class="mi">256</span><span class="p">,</span> <span class="mi">512</span><span class="p">,</span> <span class="mi">1024</span><span class="p">]</span>
|
||||
<span class="lineno">47</span> <span class="n">layers</span> <span class="o">=</span> <span class="p">[]</span>
|
||||
<span class="lineno">48</span> <span class="n">d_prev</span> <span class="o">=</span> <span class="mi">100</span>
|
||||
<span class="lineno">49</span> <span class="k">for</span> <span class="n">size</span> <span class="ow">in</span> <span class="n">layer_sizes</span><span class="p">:</span>
|
||||
<span class="lineno">50</span> <span class="n">layers</span> <span class="o">=</span> <span class="n">layers</span> <span class="o">+</span> <span class="p">[</span><span class="n">nn</span><span class="o">.</span><span class="n">Linear</span><span class="p">(</span><span class="n">d_prev</span><span class="p">,</span> <span class="n">size</span><span class="p">),</span> <span class="n">nn</span><span class="o">.</span><span class="n">LeakyReLU</span><span class="p">(</span><span class="mf">0.2</span><span class="p">)]</span>
|
||||
<span class="lineno">51</span> <span class="n">d_prev</span> <span class="o">=</span> <span class="n">size</span>
|
||||
<span class="lineno">52</span>
|
||||
<span class="lineno">53</span> <span class="bp">self</span><span class="o">.</span><span class="n">layers</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Sequential</span><span class="p">(</span><span class="o">*</span><span class="n">layers</span><span class="p">,</span> <span class="n">nn</span><span class="o">.</span><span class="n">Linear</span><span class="p">(</span><span class="n">d_prev</span><span class="p">,</span> <span class="mi">28</span> <span class="o">*</span> <span class="mi">28</span><span class="p">),</span> <span class="n">nn</span><span class="o">.</span><span class="n">Tanh</span><span class="p">())</span>
|
||||
<span class="lineno">54</span>
|
||||
<span class="lineno">55</span> <span class="bp">self</span><span class="o">.</span><span class="n">apply</span><span class="p">(</span><span class="n">weights_init</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-4'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-4'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">57</span> <span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">):</span>
|
||||
<span class="lineno">58</span> <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">layers</span><span class="p">(</span><span class="n">x</span><span class="p">)</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="n">x</span><span class="o">.</span><span class="n">shape</span><span class="p">[</span><span class="mi">0</span><span class="p">],</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">28</span><span class="p">,</span> <span class="mi">28</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-5'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-5'>#</a>
|
||||
</div>
|
||||
<h3>简单的 MLP 鉴别器</h3>
|
||||
<p>它有三个线性层，随着<code class="highlight"><span></span><span class="n">LeakyReLU</span></code>
|
||||
激活的大小逐渐减小。最后一层有一个单独的输出，它给出了输入是真实还是假的 logit。你可以通过计算它的乙状结肠来获得概率。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">61</span><span class="k">class</span> <span class="nc">Discriminator</span><span class="p">(</span><span class="n">Module</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-6'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-6'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">70</span> <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
|
||||
<span class="lineno">71</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
|
||||
<span class="lineno">72</span> <span class="n">layer_sizes</span> <span class="o">=</span> <span class="p">[</span><span class="mi">1024</span><span class="p">,</span> <span class="mi">512</span><span class="p">,</span> <span class="mi">256</span><span class="p">]</span>
|
||||
<span class="lineno">73</span> <span class="n">layers</span> <span class="o">=</span> <span class="p">[]</span>
|
||||
<span class="lineno">74</span> <span class="n">d_prev</span> <span class="o">=</span> <span class="mi">28</span> <span class="o">*</span> <span class="mi">28</span>
|
||||
<span class="lineno">75</span> <span class="k">for</span> <span class="n">size</span> <span class="ow">in</span> <span class="n">layer_sizes</span><span class="p">:</span>
|
||||
<span class="lineno">76</span> <span class="n">layers</span> <span class="o">=</span> <span class="n">layers</span> <span class="o">+</span> <span class="p">[</span><span class="n">nn</span><span class="o">.</span><span class="n">Linear</span><span class="p">(</span><span class="n">d_prev</span><span class="p">,</span> <span class="n">size</span><span class="p">),</span> <span class="n">nn</span><span class="o">.</span><span class="n">LeakyReLU</span><span class="p">(</span><span class="mf">0.2</span><span class="p">)]</span>
|
||||
<span class="lineno">77</span> <span class="n">d_prev</span> <span class="o">=</span> <span class="n">size</span>
|
||||
<span class="lineno">78</span>
|
||||
<span class="lineno">79</span> <span class="bp">self</span><span class="o">.</span><span class="n">layers</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Sequential</span><span class="p">(</span><span class="o">*</span><span class="n">layers</span><span class="p">,</span> <span class="n">nn</span><span class="o">.</span><span class="n">Linear</span><span class="p">(</span><span class="n">d_prev</span><span class="p">,</span> <span class="mi">1</span><span class="p">))</span>
|
||||
<span class="lineno">80</span> <span class="bp">self</span><span class="o">.</span><span class="n">apply</span><span class="p">(</span><span class="n">weights_init</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-7'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-7'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">82</span> <span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">):</span>
|
||||
<span class="lineno">83</span> <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">layers</span><span class="p">(</span><span class="n">x</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="n">x</span><span class="o">.</span><span class="n">shape</span><span class="p">[</span><span class="mi">0</span><span class="p">],</span> <span class="o">-</span><span class="mi">1</span><span class="p">))</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-8'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-8'>#</a>
|
||||
</div>
|
||||
<h2>配置</h2>
|
||||
<p>这扩展了 MNIST 配置以获取数据加载器以及训练和验证循环配置，从而简化了我们的实现。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">86</span><span class="k">class</span> <span class="nc">Configs</span><span class="p">(</span><span class="n">MNISTConfigs</span><span class="p">,</span> <span class="n">TrainValidConfigs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-9'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-9'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">94</span> <span class="n">device</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">device</span> <span class="o">=</span> <span class="n">DeviceConfigs</span><span class="p">()</span>
|
||||
<span class="lineno">95</span> <span class="n">dataset_transforms</span> <span class="o">=</span> <span class="s1">'mnist_gan_transforms'</span>
|
||||
<span class="lineno">96</span> <span class="n">epochs</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">10</span>
|
||||
<span class="lineno">97</span>
|
||||
<span class="lineno">98</span> <span class="n">is_save_models</span> <span class="o">=</span> <span class="kc">True</span>
|
||||
<span class="lineno">99</span> <span class="n">discriminator</span><span class="p">:</span> <span class="n">Module</span> <span class="o">=</span> <span class="s1">'mlp'</span>
|
||||
<span class="lineno">100</span> <span class="n">generator</span><span class="p">:</span> <span class="n">Module</span> <span class="o">=</span> <span class="s1">'mlp'</span>
|
||||
<span class="lineno">101</span> <span class="n">generator_optimizer</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">optim</span><span class="o">.</span><span class="n">Adam</span>
|
||||
<span class="lineno">102</span> <span class="n">discriminator_optimizer</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">optim</span><span class="o">.</span><span class="n">Adam</span>
|
||||
<span class="lineno">103</span> <span class="n">generator_loss</span><span class="p">:</span> <span class="n">GeneratorLogitsLoss</span> <span class="o">=</span> <span class="s1">'original'</span>
|
||||
<span class="lineno">104</span> <span class="n">discriminator_loss</span><span class="p">:</span> <span class="n">DiscriminatorLogitsLoss</span> <span class="o">=</span> <span class="s1">'original'</span>
|
||||
<span class="lineno">105</span> <span class="n">label_smoothing</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">0.2</span>
|
||||
<span class="lineno">106</span> <span class="n">discriminator_k</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">1</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-10'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-10'>#</a>
|
||||
</div>
|
||||
<p>初始化</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">108</span> <span class="k">def</span> <span class="nf">init</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-11'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-11'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">112</span> <span class="bp">self</span><span class="o">.</span><span class="n">state_modules</span> <span class="o">=</span> <span class="p">[]</span>
|
||||
<span class="lineno">113</span>
|
||||
<span class="lineno">114</span> <span class="n">hook_model_outputs</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">mode</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">generator</span><span class="p">,</span> <span class="s1">'generator'</span><span class="p">)</span>
|
||||
<span class="lineno">115</span> <span class="n">hook_model_outputs</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">mode</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">discriminator</span><span class="p">,</span> <span class="s1">'discriminator'</span><span class="p">)</span>
|
||||
<span class="lineno">116</span> <span class="n">tracker</span><span class="o">.</span><span class="n">set_scalar</span><span class="p">(</span><span class="s2">"loss.generator.*"</span><span class="p">,</span> <span class="kc">True</span><span class="p">)</span>
|
||||
<span class="lineno">117</span> <span class="n">tracker</span><span class="o">.</span><span class="n">set_scalar</span><span class="p">(</span><span class="s2">"loss.discriminator.*"</span><span class="p">,</span> <span class="kc">True</span><span class="p">)</span>
|
||||
<span class="lineno">118</span> <span class="n">tracker</span><span class="o">.</span><span class="n">set_image</span><span class="p">(</span><span class="s2">"generated"</span><span class="p">,</span> <span class="kc">True</span><span class="p">,</span> <span class="mi">1</span> <span class="o">/</span> <span class="mi">100</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-12'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-12'>#</a>
|
||||
</div>
|
||||
<p><span ><span class="katex-display"><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.43056em;vertical-align:0em;"></span><span class="mord mathnormal" style="margin-right:0.04398em;">z</span><span class="mspace" style="margin-right:0.2777777777777778em;"></span><span class="mrel">∼</span><span class="mspace" style="margin-right:0.2777777777777778em;"></span></span><span class="base"><span class="strut" style="height:1em;vertical-align:-0.25em;"></span><span class="mord mathnormal">p</span><span class="mopen">(</span><span class="mord mathnormal" style="margin-right:0.04398em;">z</span><span class="mclose">)</span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">120</span> <span class="k">def</span> <span class="nf">sample_z</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">batch_size</span><span class="p">:</span> <span class="nb">int</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-13'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-13'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">124</span> <span class="k">return</span> <span class="n">torch</span><span class="o">.</span><span class="n">randn</span><span class="p">(</span><span class="n">batch_size</span><span class="p">,</span> <span class="mi">100</span><span class="p">,</span> <span class="n">device</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">device</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-14'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-14'>#</a>
|
||||
</div>
|
||||
<p>迈出训练一步</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">126</span> <span class="k">def</span> <span class="nf">step</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">batch</span><span class="p">:</span> <span class="n">Any</span><span class="p">,</span> <span class="n">batch_idx</span><span class="p">:</span> <span class="n">BatchIndex</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-15'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-15'>#</a>
|
||||
</div>
|
||||
<p>设置模型状态</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">132</span> <span class="bp">self</span><span class="o">.</span><span class="n">generator</span><span class="o">.</span><span class="n">train</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">mode</span><span class="o">.</span><span class="n">is_train</span><span class="p">)</span>
|
||||
<span class="lineno">133</span> <span class="bp">self</span><span class="o">.</span><span class="n">discriminator</span><span class="o">.</span><span class="n">train</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">mode</span><span class="o">.</span><span class="n">is_train</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-16'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-16'>#</a>
|
||||
</div>
|
||||
<p>获取 MNIST 图片</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">136</span> <span class="n">data</span> <span class="o">=</span> <span class="n">batch</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">device</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-17'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-17'>#</a>
|
||||
</div>
|
||||
<p>在训练模式中增加步数</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">139</span> <span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">mode</span><span class="o">.</span><span class="n">is_train</span><span class="p">:</span>
|
||||
<span class="lineno">140</span> <span class="n">tracker</span><span class="o">.</span><span class="n">add_global_step</span><span class="p">(</span><span class="nb">len</span><span class="p">(</span><span class="n">data</span><span class="p">))</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-18'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-18'>#</a>
|
||||
</div>
|
||||
<p>训练鉴别器</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">143</span> <span class="k">with</span> <span class="n">monit</span><span class="o">.</span><span class="n">section</span><span class="p">(</span><span class="s2">"discriminator"</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-19'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-19'>#</a>
|
||||
</div>
|
||||
<p>获得鉴别器损失</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">145</span> <span class="n">loss</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">calc_discriminator_loss</span><span class="p">(</span><span class="n">data</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-20'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-20'>#</a>
|
||||
</div>
|
||||
<p>火车</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">148</span> <span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">mode</span><span class="o">.</span><span class="n">is_train</span><span class="p">:</span>
|
||||
<span class="lineno">149</span> <span class="bp">self</span><span class="o">.</span><span class="n">discriminator_optimizer</span><span class="o">.</span><span class="n">zero_grad</span><span class="p">()</span>
|
||||
<span class="lineno">150</span> <span class="n">loss</span><span class="o">.</span><span class="n">backward</span><span class="p">()</span>
|
||||
<span class="lineno">151</span> <span class="k">if</span> <span class="n">batch_idx</span><span class="o">.</span><span class="n">is_last</span><span class="p">:</span>
|
||||
<span class="lineno">152</span> <span class="n">tracker</span><span class="o">.</span><span class="n">add</span><span class="p">(</span><span class="s1">'discriminator'</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">discriminator</span><span class="p">)</span>
|
||||
<span class="lineno">153</span> <span class="bp">self</span><span class="o">.</span><span class="n">discriminator_optimizer</span><span class="o">.</span><span class="n">step</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-21'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-21'>#</a>
|
||||
</div>
|
||||
<p>每隔一次训练发电机<code class="highlight"><span></span><span class="n">discriminator_k</span></code>
|
||||
</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">156</span> <span class="k">if</span> <span class="n">batch_idx</span><span class="o">.</span><span class="n">is_interval</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">discriminator_k</span><span class="p">):</span>
|
||||
<span class="lineno">157</span> <span class="k">with</span> <span class="n">monit</span><span class="o">.</span><span class="n">section</span><span class="p">(</span><span class="s2">"generator"</span><span class="p">):</span>
|
||||
<span class="lineno">158</span> <span class="n">loss</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">calc_generator_loss</span><span class="p">(</span><span class="n">data</span><span class="o">.</span><span class="n">shape</span><span class="p">[</span><span class="mi">0</span><span class="p">])</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-22'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-22'>#</a>
|
||||
</div>
|
||||
<p>火车</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">161</span> <span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">mode</span><span class="o">.</span><span class="n">is_train</span><span class="p">:</span>
|
||||
<span class="lineno">162</span> <span class="bp">self</span><span class="o">.</span><span class="n">generator_optimizer</span><span class="o">.</span><span class="n">zero_grad</span><span class="p">()</span>
|
||||
<span class="lineno">163</span> <span class="n">loss</span><span class="o">.</span><span class="n">backward</span><span class="p">()</span>
|
||||
<span class="lineno">164</span> <span class="k">if</span> <span class="n">batch_idx</span><span class="o">.</span><span class="n">is_last</span><span class="p">:</span>
|
||||
<span class="lineno">165</span> <span class="n">tracker</span><span class="o">.</span><span class="n">add</span><span class="p">(</span><span class="s1">'generator'</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">generator</span><span class="p">)</span>
|
||||
<span class="lineno">166</span> <span class="bp">self</span><span class="o">.</span><span class="n">generator_optimizer</span><span class="o">.</span><span class="n">step</span><span class="p">()</span>
|
||||
<span class="lineno">167</span>
|
||||
<span class="lineno">168</span> <span class="n">tracker</span><span class="o">.</span><span class="n">save</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-23'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-23'>#</a>
|
||||
</div>
|
||||
<p>计算鉴别器损失</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">170</span> <span class="k">def</span> <span class="nf">calc_discriminator_loss</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">data</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-24'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-24'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">174</span> <span class="n">latent</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">sample_z</span><span class="p">(</span><span class="n">data</span><span class="o">.</span><span class="n">shape</span><span class="p">[</span><span class="mi">0</span><span class="p">])</span>
|
||||
<span class="lineno">175</span> <span class="n">logits_true</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">discriminator</span><span class="p">(</span><span class="n">data</span><span class="p">)</span>
|
||||
<span class="lineno">176</span> <span class="n">logits_false</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">discriminator</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">generator</span><span class="p">(</span><span class="n">latent</span><span class="p">)</span><span class="o">.</span><span class="n">detach</span><span class="p">())</span>
|
||||
<span class="lineno">177</span> <span class="n">loss_true</span><span class="p">,</span> <span class="n">loss_false</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">discriminator_loss</span><span class="p">(</span><span class="n">logits_true</span><span class="p">,</span> <span class="n">logits_false</span><span class="p">)</span>
|
||||
<span class="lineno">178</span> <span class="n">loss</span> <span class="o">=</span> <span class="n">loss_true</span> <span class="o">+</span> <span class="n">loss_false</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-25'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-25'>#</a>
|
||||
</div>
|
||||
<p>日志的东西</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">181</span> <span class="n">tracker</span><span class="o">.</span><span class="n">add</span><span class="p">(</span><span class="s2">"loss.discriminator.true."</span><span class="p">,</span> <span class="n">loss_true</span><span class="p">)</span>
|
||||
<span class="lineno">182</span> <span class="n">tracker</span><span class="o">.</span><span class="n">add</span><span class="p">(</span><span class="s2">"loss.discriminator.false."</span><span class="p">,</span> <span class="n">loss_false</span><span class="p">)</span>
|
||||
<span class="lineno">183</span> <span class="n">tracker</span><span class="o">.</span><span class="n">add</span><span class="p">(</span><span class="s2">"loss.discriminator."</span><span class="p">,</span> <span class="n">loss</span><span class="p">)</span>
|
||||
<span class="lineno">184</span>
|
||||
<span class="lineno">185</span> <span class="k">return</span> <span class="n">loss</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-26'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-26'>#</a>
|
||||
</div>
|
||||
<p>计算发电机损耗</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">187</span> <span class="k">def</span> <span class="nf">calc_generator_loss</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">batch_size</span><span class="p">:</span> <span class="nb">int</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-27'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-27'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">191</span> <span class="n">latent</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">sample_z</span><span class="p">(</span><span class="n">batch_size</span><span class="p">)</span>
|
||||
<span class="lineno">192</span> <span class="n">generated_images</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">generator</span><span class="p">(</span><span class="n">latent</span><span class="p">)</span>
|
||||
<span class="lineno">193</span> <span class="n">logits</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">discriminator</span><span class="p">(</span><span class="n">generated_images</span><span class="p">)</span>
|
||||
<span class="lineno">194</span> <span class="n">loss</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">generator_loss</span><span class="p">(</span><span class="n">logits</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-28'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-28'>#</a>
|
||||
</div>
|
||||
<p>日志的东西</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">197</span> <span class="n">tracker</span><span class="o">.</span><span class="n">add</span><span class="p">(</span><span class="s1">'generated'</span><span class="p">,</span> <span class="n">generated_images</span><span class="p">[</span><span class="mi">0</span><span class="p">:</span><span class="mi">6</span><span class="p">])</span>
|
||||
<span class="lineno">198</span> <span class="n">tracker</span><span class="o">.</span><span class="n">add</span><span class="p">(</span><span class="s2">"loss.generator."</span><span class="p">,</span> <span class="n">loss</span><span class="p">)</span>
|
||||
<span class="lineno">199</span>
|
||||
<span class="lineno">200</span> <span class="k">return</span> <span class="n">loss</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-29'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-29'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">205</span><span class="nd">@option</span><span class="p">(</span><span class="n">Configs</span><span class="o">.</span><span class="n">dataset_transforms</span><span class="p">)</span>
|
||||
<span class="lineno">206</span><span class="k">def</span> <span class="nf">mnist_gan_transforms</span><span class="p">():</span>
|
||||
<span class="lineno">207</span> <span class="k">return</span> <span class="n">transforms</span><span class="o">.</span><span class="n">Compose</span><span class="p">([</span>
|
||||
<span class="lineno">208</span> <span class="n">transforms</span><span class="o">.</span><span class="n">ToTensor</span><span class="p">(),</span>
|
||||
<span class="lineno">209</span> <span class="n">transforms</span><span class="o">.</span><span class="n">Normalize</span><span class="p">((</span><span class="mf">0.5</span><span class="p">,),</span> <span class="p">(</span><span class="mf">0.5</span><span class="p">,))</span>
|
||||
<span class="lineno">210</span> <span class="p">])</span>
|
||||
<span class="lineno">211</span>
|
||||
<span class="lineno">212</span>
|
||||
<span class="lineno">213</span><span class="nd">@option</span><span class="p">(</span><span class="n">Configs</span><span class="o">.</span><span class="n">discriminator_optimizer</span><span class="p">)</span>
|
||||
<span class="lineno">214</span><span class="k">def</span> <span class="nf">_discriminator_optimizer</span><span class="p">(</span><span class="n">c</span><span class="p">:</span> <span class="n">Configs</span><span class="p">):</span>
|
||||
<span class="lineno">215</span> <span class="n">opt_conf</span> <span class="o">=</span> <span class="n">OptimizerConfigs</span><span class="p">()</span>
|
||||
<span class="lineno">216</span> <span class="n">opt_conf</span><span class="o">.</span><span class="n">optimizer</span> <span class="o">=</span> <span class="s1">'Adam'</span>
|
||||
<span class="lineno">217</span> <span class="n">opt_conf</span><span class="o">.</span><span class="n">parameters</span> <span class="o">=</span> <span class="n">c</span><span class="o">.</span><span class="n">discriminator</span><span class="o">.</span><span class="n">parameters</span><span class="p">()</span>
|
||||
<span class="lineno">218</span> <span class="n">opt_conf</span><span class="o">.</span><span class="n">learning_rate</span> <span class="o">=</span> <span class="mf">2.5e-4</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-30'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-30'>#</a>
|
||||
</div>
|
||||
<p>设置梯度第一时刻的指数衰减率<code class="highlight"><span></span><span class="mf">0.5</span></code>
|
||||
非常重要。<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.8888799999999999em;vertical-align:-0.19444em;"></span><span class="mord coloredeq eqb" style=""><span class="mord" style=""><span class="mord mathnormal" style="margin-right:0.05278em">β</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.30110799999999993em;"><span style="top:-2.5500000000000003em;margin-left:-0.05278em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style="">1</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span></span></span></span></span></span>默认为<code class="highlight"><span></span><span class="mf">0.9</span></code>
|
||||
失败。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">222</span> <span class="n">opt_conf</span><span class="o">.</span><span class="n">betas</span> <span class="o">=</span> <span class="p">(</span><span class="mf">0.5</span><span class="p">,</span> <span class="mf">0.999</span><span class="p">)</span>
|
||||
<span class="lineno">223</span> <span class="k">return</span> <span class="n">opt_conf</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-31'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-31'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">226</span><span class="nd">@option</span><span class="p">(</span><span class="n">Configs</span><span class="o">.</span><span class="n">generator_optimizer</span><span class="p">)</span>
|
||||
<span class="lineno">227</span><span class="k">def</span> <span class="nf">_generator_optimizer</span><span class="p">(</span><span class="n">c</span><span class="p">:</span> <span class="n">Configs</span><span class="p">):</span>
|
||||
<span class="lineno">228</span> <span class="n">opt_conf</span> <span class="o">=</span> <span class="n">OptimizerConfigs</span><span class="p">()</span>
|
||||
<span class="lineno">229</span> <span class="n">opt_conf</span><span class="o">.</span><span class="n">optimizer</span> <span class="o">=</span> <span class="s1">'Adam'</span>
|
||||
<span class="lineno">230</span> <span class="n">opt_conf</span><span class="o">.</span><span class="n">parameters</span> <span class="o">=</span> <span class="n">c</span><span class="o">.</span><span class="n">generator</span><span class="o">.</span><span class="n">parameters</span><span class="p">()</span>
|
||||
<span class="lineno">231</span> <span class="n">opt_conf</span><span class="o">.</span><span class="n">learning_rate</span> <span class="o">=</span> <span class="mf">2.5e-4</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-32'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-32'>#</a>
|
||||
</div>
|
||||
<p>设置梯度第一时刻的指数衰减率<code class="highlight"><span></span><span class="mf">0.5</span></code>
|
||||
非常重要。<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.8888799999999999em;vertical-align:-0.19444em;"></span><span class="mord coloredeq eqb" style=""><span class="mord" style=""><span class="mord mathnormal" style="margin-right:0.05278em">β</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.30110799999999993em;"><span style="top:-2.5500000000000003em;margin-left:-0.05278em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight" style=""><span class="mord mtight" style="">1</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span></span></span></span></span></span>默认为<code class="highlight"><span></span><span class="mf">0.9</span></code>
|
||||
失败。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">235</span> <span class="n">opt_conf</span><span class="o">.</span><span class="n">betas</span> <span class="o">=</span> <span class="p">(</span><span class="mf">0.5</span><span class="p">,</span> <span class="mf">0.999</span><span class="p">)</span>
|
||||
<span class="lineno">236</span> <span class="k">return</span> <span class="n">opt_conf</span>
|
||||
<span class="lineno">237</span>
|
||||
<span class="lineno">238</span>
|
||||
<span class="lineno">239</span><span class="n">calculate</span><span class="p">(</span><span class="n">Configs</span><span class="o">.</span><span class="n">generator</span><span class="p">,</span> <span class="s1">'mlp'</span><span class="p">,</span> <span class="k">lambda</span> <span class="n">c</span><span class="p">:</span> <span class="n">Generator</span><span class="p">()</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">device</span><span class="p">))</span>
|
||||
<span class="lineno">240</span><span class="n">calculate</span><span class="p">(</span><span class="n">Configs</span><span class="o">.</span><span class="n">discriminator</span><span class="p">,</span> <span class="s1">'mlp'</span><span class="p">,</span> <span class="k">lambda</span> <span class="n">c</span><span class="p">:</span> <span class="n">Discriminator</span><span class="p">()</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">device</span><span class="p">))</span>
|
||||
<span class="lineno">241</span><span class="n">calculate</span><span class="p">(</span><span class="n">Configs</span><span class="o">.</span><span class="n">generator_loss</span><span class="p">,</span> <span class="s1">'original'</span><span class="p">,</span> <span class="k">lambda</span> <span class="n">c</span><span class="p">:</span> <span class="n">GeneratorLogitsLoss</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">label_smoothing</span><span class="p">)</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">device</span><span class="p">))</span>
|
||||
<span class="lineno">242</span><span class="n">calculate</span><span class="p">(</span><span class="n">Configs</span><span class="o">.</span><span class="n">discriminator_loss</span><span class="p">,</span> <span class="s1">'original'</span><span class="p">,</span> <span class="k">lambda</span> <span class="n">c</span><span class="p">:</span> <span class="n">DiscriminatorLogitsLoss</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">label_smoothing</span><span class="p">)</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">device</span><span class="p">))</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-33'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-33'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">245</span><span class="k">def</span> <span class="nf">main</span><span class="p">():</span>
|
||||
<span class="lineno">246</span> <span class="n">conf</span> <span class="o">=</span> <span class="n">Configs</span><span class="p">()</span>
|
||||
<span class="lineno">247</span> <span class="n">experiment</span><span class="o">.</span><span class="n">create</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">'mnist_gan'</span><span class="p">,</span> <span class="n">comment</span><span class="o">=</span><span class="s1">'test'</span><span class="p">)</span>
|
||||
<span class="lineno">248</span> <span class="n">experiment</span><span class="o">.</span><span class="n">configs</span><span class="p">(</span><span class="n">conf</span><span class="p">,</span>
|
||||
<span class="lineno">249</span> <span class="p">{</span><span class="s1">'label_smoothing'</span><span class="p">:</span> <span class="mf">0.01</span><span class="p">})</span>
|
||||
<span class="lineno">250</span> <span class="k">with</span> <span class="n">experiment</span><span class="o">.</span><span class="n">start</span><span class="p">():</span>
|
||||
<span class="lineno">251</span> <span class="n">conf</span><span class="o">.</span><span class="n">run</span><span class="p">()</span>
|
||||
<span class="lineno">252</span>
|
||||
<span class="lineno">253</span>
|
||||
<span class="lineno">254</span><span class="k">if</span> <span class="vm">__name__</span> <span class="o">==</span> <span class="s1">'__main__'</span><span class="p">:</span>
|
||||
<span class="lineno">255</span> <span class="n">main</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
File diff suppressed because one or more lines are too long
@@ -0,0 +1,127 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content=""/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="生成式对抗网络-GAN"/>
|
||||
<meta name="twitter:description" content=""/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/gan/original/readme.html"/>
|
||||
<meta property="og:title" content="生成式对抗网络-GAN"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="生成式对抗网络-GAN"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="生成式对抗网络-GAN"/>
|
||||
<meta property="og:description" content=""/>
|
||||
|
||||
<title>生成式对抗网络-GAN</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/gan/original/readme.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="../index.html">gan</a>
|
||||
<a class="parent" href="index.html">original</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/gan/original/readme.md" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1><a href="https://nn.labml.ai/gan/original/index.html">生成式对抗网络-GAN</a></h1>
|
||||
<p>这是<a href="https://arxiv.org/abs/1406.2661">生成对抗网络</a>的带注释的实现。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
File diff suppressed because one or more lines are too long
|
After Width: | Height: | Size: 9.0 KiB |
File diff suppressed because one or more lines are too long
Binary file not shown.
|
After Width: | Height: | Size: 474 KiB |
File diff suppressed because one or more lines are too long
|
After Width: | Height: | Size: 24 KiB |
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
|
After Width: | Height: | Size: 9.0 KiB |
File diff suppressed because one or more lines are too long
|
After Width: | Height: | Size: 42 KiB |
@@ -0,0 +1,127 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content=""/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="StyleGan 2"/>
|
||||
<meta name="twitter:description" content=""/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/gan/stylegan/readme.html"/>
|
||||
<meta property="og:title" content="StyleGan 2"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="StyleGan 2"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="StyleGan 2"/>
|
||||
<meta property="og:description" content=""/>
|
||||
|
||||
<title>StyleGan 2</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/gan/stylegan/readme.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="../index.html">gan</a>
|
||||
<a class="parent" href="index.html">stylegan</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/gan/stylegan/readme.md" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1><a href="https://nn.labml.ai/gan/stylegan/index.html">StyleGan 2</a></h1>
|
||||
<p>这是《<a href="https://arxiv.org/abs/1912.04958">分析和提高 StyleGan 的图像质量》</a><a href="https://pytorch.org">一文的 PyTorch</a> 实现，该论文介绍了 <strong>StyleGan2</strong>。StyleGan 2 是对论文《生成<a href="https://arxiv.org/abs/1812.04948">对抗网络的基于样式的生成器架构》中对</a> <strong>StyleG</strong> an 的改进。StyleG <strong>an 基于论文《逐步</strong><a href="https://arxiv.org/abs/1710.10196">生长 GaN 以提高质量、稳定性和变异性》中的渐进式 GAN</a>。这三篇论文均出自 <a href="https://twitter.com/NVIDIAAI">NVIDIA AI</a> 的同一位作者。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
File diff suppressed because one or more lines are too long
|
After Width: | Height: | Size: 12 KiB |
File diff suppressed because one or more lines are too long
|
After Width: | Height: | Size: 43 KiB |
File diff suppressed because one or more lines are too long
|
After Width: | Height: | Size: 39 KiB |
File diff suppressed because one or more lines are too long
|
After Width: | Height: | Size: 26 KiB |
File diff suppressed because one or more lines are too long
|
After Width: | Height: | Size: 9.5 KiB |
@@ -0,0 +1,236 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content="本实验使用卷积神经网络生成 MNIST 图像。"/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="WGAN 使用 MNIST 进行实验"/>
|
||||
<meta name="twitter:description" content="本实验使用卷积神经网络生成 MNIST 图像。"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/gan/wasserstein/experiment.html"/>
|
||||
<meta property="og:title" content="WGAN 使用 MNIST 进行实验"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="WGAN 使用 MNIST 进行实验"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="WGAN 使用 MNIST 进行实验"/>
|
||||
<meta property="og:description" content="本实验使用卷积神经网络生成 MNIST 图像。"/>
|
||||
|
||||
<title>WGAN 使用 MNIST 进行实验</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/gan/wasserstein/experiment.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="../index.html">gan</a>
|
||||
<a class="parent" href="index.html">wasserstein</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/gan/wasserstein/experiment.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1>WGAN 使用 MNIST 进行实验</h1>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">9</span><span></span><span class="kn">from</span> <span class="nn">labml</span> <span class="kn">import</span> <span class="n">experiment</span>
|
||||
<span class="lineno">10</span>
|
||||
<span class="lineno">11</span><span class="kn">from</span> <span class="nn">labml.configs</span> <span class="kn">import</span> <span class="n">calculate</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-1'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-1'>#</a>
|
||||
</div>
|
||||
<p>从 <a href="../dcgan/index.html">DCGAN 实验</a>导入配置</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">13</span><span class="kn">from</span> <span class="nn">labml_nn.gan.dcgan</span> <span class="kn">import</span> <span class="n">Configs</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-2'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-2'>#</a>
|
||||
</div>
|
||||
<p>进口 <a href="./index.html">Wasserstein GAN 亏损</a></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">16</span><span class="kn">from</span> <span class="nn">labml_nn.gan.wasserstein</span> <span class="kn">import</span> <span class="n">GeneratorLoss</span><span class="p">,</span> <span class="n">DiscriminatorLoss</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-3'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-3'>#</a>
|
||||
</div>
|
||||
<p>为 Wasserstein GAN 损耗设置配置选项</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">19</span><span class="n">calculate</span><span class="p">(</span><span class="n">Configs</span><span class="o">.</span><span class="n">generator_loss</span><span class="p">,</span> <span class="s1">'wasserstein'</span><span class="p">,</span> <span class="k">lambda</span> <span class="n">c</span><span class="p">:</span> <span class="n">GeneratorLoss</span><span class="p">())</span>
|
||||
<span class="lineno">20</span><span class="n">calculate</span><span class="p">(</span><span class="n">Configs</span><span class="o">.</span><span class="n">discriminator_loss</span><span class="p">,</span> <span class="s1">'wasserstein'</span><span class="p">,</span> <span class="k">lambda</span> <span class="n">c</span><span class="p">:</span> <span class="n">DiscriminatorLoss</span><span class="p">())</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-4'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-4'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">23</span><span class="k">def</span> <span class="nf">main</span><span class="p">():</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-5'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-5'>#</a>
|
||||
</div>
|
||||
<p>创建配置对象</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">25</span> <span class="n">conf</span> <span class="o">=</span> <span class="n">Configs</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-6'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-6'>#</a>
|
||||
</div>
|
||||
<p>创建实验</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">27</span> <span class="n">experiment</span><span class="o">.</span><span class="n">create</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">'mnist_wassertein_dcgan'</span><span class="p">,</span> <span class="n">comment</span><span class="o">=</span><span class="s1">'test'</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-7'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-7'>#</a>
|
||||
</div>
|
||||
<p>覆盖配置</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">29</span> <span class="n">experiment</span><span class="o">.</span><span class="n">configs</span><span class="p">(</span><span class="n">conf</span><span class="p">,</span>
|
||||
<span class="lineno">30</span> <span class="p">{</span>
|
||||
<span class="lineno">31</span> <span class="s1">'discriminator'</span><span class="p">:</span> <span class="s1">'cnn'</span><span class="p">,</span>
|
||||
<span class="lineno">32</span> <span class="s1">'generator'</span><span class="p">:</span> <span class="s1">'cnn'</span><span class="p">,</span>
|
||||
<span class="lineno">33</span> <span class="s1">'label_smoothing'</span><span class="p">:</span> <span class="mf">0.01</span><span class="p">,</span>
|
||||
<span class="lineno">34</span> <span class="s1">'generator_loss'</span><span class="p">:</span> <span class="s1">'wasserstein'</span><span class="p">,</span>
|
||||
<span class="lineno">35</span> <span class="s1">'discriminator_loss'</span><span class="p">:</span> <span class="s1">'wasserstein'</span><span class="p">,</span>
|
||||
<span class="lineno">36</span> <span class="p">})</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-8'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-8'>#</a>
|
||||
</div>
|
||||
<p>开始实验并运行训练循环</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">39</span> <span class="k">with</span> <span class="n">experiment</span><span class="o">.</span><span class="n">start</span><span class="p">():</span>
|
||||
<span class="lineno">40</span> <span class="n">conf</span><span class="o">.</span><span class="n">run</span><span class="p">()</span>
|
||||
<span class="lineno">41</span>
|
||||
<span class="lineno">42</span>
|
||||
<span class="lineno">43</span><span class="k">if</span> <span class="vm">__name__</span> <span class="o">==</span> <span class="s1">'__main__'</span><span class="p">:</span>
|
||||
<span class="lineno">44</span> <span class="n">main</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
@@ -0,0 +1,378 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content="本实验使用卷积神经网络生成 MNIST 图像。"/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="WGAN-GP 使用 MNIST 进行实验"/>
|
||||
<meta name="twitter:description" content="本实验使用卷积神经网络生成 MNIST 图像。"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/gan/wasserstein/gradient_penalty/experiment.html"/>
|
||||
<meta property="og:title" content="WGAN-GP 使用 MNIST 进行实验"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="WGAN-GP 使用 MNIST 进行实验"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="WGAN-GP 使用 MNIST 进行实验"/>
|
||||
<meta property="og:description" content="本实验使用卷积神经网络生成 MNIST 图像。"/>
|
||||
|
||||
<title>WGAN-GP 使用 MNIST 进行实验</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../../../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/gan/wasserstein/gradient_penalty/experiment.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="../../index.html">gan</a>
|
||||
<a class="parent" href="../index.html">wasserstein</a>
|
||||
<a class="parent" href="index.html">gradient_penalty</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/gan/wasserstein/gradient_penalty/experiment.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1>WGAN-GP 使用 MNIST 进行实验</h1>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">10</span><span></span><span class="kn">import</span> <span class="nn">torch</span>
|
||||
<span class="lineno">11</span>
|
||||
<span class="lineno">12</span><span class="kn">from</span> <span class="nn">labml</span> <span class="kn">import</span> <span class="n">experiment</span><span class="p">,</span> <span class="n">tracker</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-1'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-1'>#</a>
|
||||
</div>
|
||||
<p>从 <a href="../experiment.html">Wasserstein 实验</a>导入配置</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">14</span><span class="kn">from</span> <span class="nn">labml_nn.gan.wasserstein.experiment</span> <span class="kn">import</span> <span class="n">Configs</span> <span class="k">as</span> <span class="n">OriginalConfigs</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-2'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-2'>#</a>
|
||||
</div>
|
||||
<p></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">16</span><span class="kn">from</span> <span class="nn">labml_nn.gan.wasserstein.gradient_penalty</span> <span class="kn">import</span> <span class="n">GradientPenalty</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-3'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-3'>#</a>
|
||||
</div>
|
||||
<h2>配置类</h2>
|
||||
<p>我们扩展了<a href="../../original/experiment.html">最初的 GAN 实现</a>，并覆盖了鉴别器（批评者）损失计算，以包括梯度惩罚。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">19</span><span class="k">class</span> <span class="nc">Configs</span><span class="p">(</span><span class="n">OriginalConfigs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-4'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-4'>#</a>
|
||||
</div>
|
||||
<p>梯度惩罚系数<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.69444em;vertical-align:0em;"></span><span class="mord mathnormal">λ</span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">28</span> <span class="n">gradient_penalty_coefficient</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">10.0</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-5'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-5'>#</a>
|
||||
</div>
|
||||
<p></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">30</span> <span class="n">gradient_penalty</span> <span class="o">=</span> <span class="n">GradientPenalty</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-6'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-6'>#</a>
|
||||
</div>
|
||||
<p>这会覆盖最初的鉴别器损耗计算，并包括梯度惩罚。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">32</span> <span class="k">def</span> <span class="nf">calc_discriminator_loss</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">data</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-7'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-7'>#</a>
|
||||
</div>
|
||||
<p>需要开启梯度<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.43056em;vertical-align:0em;"></span><span class="mord coloredeq eqe" style=""><span class="mord mathnormal" style="">x</span></span></span></span></span></span>才能计算梯度损失</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">38</span> <span class="n">data</span><span class="o">.</span><span class="n">requires_grad_</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-8'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-8'>#</a>
|
||||
</div>
|
||||
<p>样本<span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.43056em;vertical-align:0em;"></span><span class="mord mathnormal" style="margin-right:0.04398em;">z</span><span class="mspace" style="margin-right:0.2777777777777778em;"></span><span class="mrel">∼</span><span class="mspace" style="margin-right:0.2777777777777778em;"></span></span><span class="base"><span class="strut" style="height:1em;vertical-align:-0.25em;"></span><span class="mord mathnormal">p</span><span class="mopen">(</span><span class="mord mathnormal" style="margin-right:0.04398em;">z</span><span class="mclose">)</span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">40</span> <span class="n">latent</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">sample_z</span><span class="p">(</span><span class="n">data</span><span class="o">.</span><span class="n">shape</span><span class="p">[</span><span class="mi">0</span><span class="p">])</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-9'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-9'>#</a>
|
||||
</div>
|
||||
<p><span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:1em;vertical-align:-0.25em;"></span><span class="mord mathnormal" style="margin-right:0.02778em;">D</span><span class="mopen">(</span><span class="mord coloredeq eqe" style=""><span class="mord mathnormal" style="">x</span></span><span class="mclose">)</span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">42</span> <span class="n">f_real</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">discriminator</span><span class="p">(</span><span class="n">data</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-10'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-10'>#</a>
|
||||
</div>
|
||||
<p><span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:1em;vertical-align:-0.25em;"></span><span class="mord mathnormal" style="margin-right:0.02778em;">D</span><span class="mopen">(</span><span class="mord"><span class="mord mathnormal">G</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.33610799999999996em;"><span style="top:-2.5500000000000003em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight"><span class="mord mathnormal mtight" style="margin-right:0.02778em;">θ</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span><span class="mopen">(</span><span class="mord mathnormal" style="margin-right:0.04398em;">z</span><span class="mclose">))</span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">44</span> <span class="n">f_fake</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">discriminator</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">generator</span><span class="p">(</span><span class="n">latent</span><span class="p">)</span><span class="o">.</span><span class="n">detach</span><span class="p">())</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-11'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-11'>#</a>
|
||||
</div>
|
||||
<p>获得鉴别器损失</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">46</span> <span class="n">loss_true</span><span class="p">,</span> <span class="n">loss_false</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">discriminator_loss</span><span class="p">(</span><span class="n">f_real</span><span class="p">,</span> <span class="n">f_fake</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-12'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-12'>#</a>
|
||||
</div>
|
||||
<p>计算训练模式中的梯度惩罚</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">48</span> <span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">mode</span><span class="o">.</span><span class="n">is_train</span><span class="p">:</span>
|
||||
<span class="lineno">49</span> <span class="n">gradient_penalty</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">gradient_penalty</span><span class="p">(</span><span class="n">data</span><span class="p">,</span> <span class="n">f_real</span><span class="p">)</span>
|
||||
<span class="lineno">50</span> <span class="n">tracker</span><span class="o">.</span><span class="n">add</span><span class="p">(</span><span class="s2">"loss.gp."</span><span class="p">,</span> <span class="n">gradient_penalty</span><span class="p">)</span>
|
||||
<span class="lineno">51</span> <span class="n">loss</span> <span class="o">=</span> <span class="n">loss_true</span> <span class="o">+</span> <span class="n">loss_false</span> <span class="o">+</span> <span class="bp">self</span><span class="o">.</span><span class="n">gradient_penalty_coefficient</span> <span class="o">*</span> <span class="n">gradient_penalty</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-13'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-13'>#</a>
|
||||
</div>
|
||||
<p>否则跳过梯度惩罚</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">53</span> <span class="k">else</span><span class="p">:</span>
|
||||
<span class="lineno">54</span> <span class="n">loss</span> <span class="o">=</span> <span class="n">loss_true</span> <span class="o">+</span> <span class="n">loss_false</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-14'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-14'>#</a>
|
||||
</div>
|
||||
<p>日志的东西</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">57</span> <span class="n">tracker</span><span class="o">.</span><span class="n">add</span><span class="p">(</span><span class="s2">"loss.discriminator.true."</span><span class="p">,</span> <span class="n">loss_true</span><span class="p">)</span>
|
||||
<span class="lineno">58</span> <span class="n">tracker</span><span class="o">.</span><span class="n">add</span><span class="p">(</span><span class="s2">"loss.discriminator.false."</span><span class="p">,</span> <span class="n">loss_false</span><span class="p">)</span>
|
||||
<span class="lineno">59</span> <span class="n">tracker</span><span class="o">.</span><span class="n">add</span><span class="p">(</span><span class="s2">"loss.discriminator."</span><span class="p">,</span> <span class="n">loss</span><span class="p">)</span>
|
||||
<span class="lineno">60</span>
|
||||
<span class="lineno">61</span> <span class="k">return</span> <span class="n">loss</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-15'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-15'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">64</span><span class="k">def</span> <span class="nf">main</span><span class="p">():</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-16'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-16'>#</a>
|
||||
</div>
|
||||
<p>创建配置对象</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">66</span> <span class="n">conf</span> <span class="o">=</span> <span class="n">Configs</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-17'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-17'>#</a>
|
||||
</div>
|
||||
<p>创建实验</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">68</span> <span class="n">experiment</span><span class="o">.</span><span class="n">create</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">'mnist_wassertein_gp_dcgan'</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-18'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-18'>#</a>
|
||||
</div>
|
||||
<p>覆盖配置</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">70</span> <span class="n">experiment</span><span class="o">.</span><span class="n">configs</span><span class="p">(</span><span class="n">conf</span><span class="p">,</span>
|
||||
<span class="lineno">71</span> <span class="p">{</span>
|
||||
<span class="lineno">72</span> <span class="s1">'discriminator'</span><span class="p">:</span> <span class="s1">'cnn'</span><span class="p">,</span>
|
||||
<span class="lineno">73</span> <span class="s1">'generator'</span><span class="p">:</span> <span class="s1">'cnn'</span><span class="p">,</span>
|
||||
<span class="lineno">74</span> <span class="s1">'label_smoothing'</span><span class="p">:</span> <span class="mf">0.01</span><span class="p">,</span>
|
||||
<span class="lineno">75</span> <span class="s1">'generator_loss'</span><span class="p">:</span> <span class="s1">'wasserstein'</span><span class="p">,</span>
|
||||
<span class="lineno">76</span> <span class="s1">'discriminator_loss'</span><span class="p">:</span> <span class="s1">'wasserstein'</span><span class="p">,</span>
|
||||
<span class="lineno">77</span> <span class="s1">'discriminator_k'</span><span class="p">:</span> <span class="mi">5</span><span class="p">,</span>
|
||||
<span class="lineno">78</span> <span class="p">})</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-19'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-19'>#</a>
|
||||
</div>
|
||||
<p>开始实验并运行训练循环</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">81</span> <span class="k">with</span> <span class="n">experiment</span><span class="o">.</span><span class="n">start</span><span class="p">():</span>
|
||||
<span class="lineno">82</span> <span class="n">conf</span><span class="o">.</span><span class="n">run</span><span class="p">()</span>
|
||||
<span class="lineno">83</span>
|
||||
<span class="lineno">84</span>
|
||||
<span class="lineno">85</span><span class="k">if</span> <span class="vm">__name__</span> <span class="o">==</span> <span class="s1">'__main__'</span><span class="p">:</span>
|
||||
<span class="lineno">86</span> <span class="n">main</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../../../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
File diff suppressed because one or more lines are too long
@@ -0,0 +1,131 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content=""/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="Wasserstein GAN (WGAN-GP) 的梯度惩罚"/>
|
||||
<meta name="twitter:description" content=""/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/gan/wasserstein/gradient_penalty/readme.html"/>
|
||||
<meta property="og:title" content="Wasserstein GAN (WGAN-GP) 的梯度惩罚"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="Wasserstein GAN (WGAN-GP) 的梯度惩罚"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="Wasserstein GAN (WGAN-GP) 的梯度惩罚"/>
|
||||
<meta property="og:description" content=""/>
|
||||
|
||||
<title>Wasserstein GAN (WGAN-GP) 的梯度惩罚</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../../../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/gan/wasserstein/gradient_penalty/readme.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="../../index.html">gan</a>
|
||||
<a class="parent" href="../index.html">wasserstein</a>
|
||||
<a class="parent" href="index.html">gradient_penalty</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/gan/wasserstein/gradient_penalty/readme.md" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1><a href="https://nn.labml.ai/gan/wasserstein/gradient_penalty/index.html">Wasserstein GAN (WGAN-GP) 的梯度惩罚</a></h1>
|
||||
<p>这是 <a href="https://arxiv.org/abs/1704.00028">Wasserstein GAN 改进训练的</a>实现。</p>
|
||||
<p><a href="https://nn.labml.ai/gan/wasserstein/index.html">WGAN</a> 建议削减权重以对鉴别器网络强制执行 Lipschitz 限制（评论家）。这个和其他权重约束，如L2标准削减、权重标准化、L1、L2权重衰减都有问题：</p>
|
||||
<p>1.限制鉴别器的容量 2.分解和消失渐变（不带<a href="https://nn.labml.ai/normalization/batch_norm/index.html">批量归一化</a>）。</p>
|
||||
<p>论文《<a href="https://arxiv.org/abs/1704.00028">改进了 Wasserstein GaN 的训练》</a>提出了改进 Lipschitz 约束的更好方法，即梯度惩罚。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../../../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
File diff suppressed because one or more lines are too long
@@ -0,0 +1,127 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content=""/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="Wasserstein GAN-WGAN"/>
|
||||
<meta name="twitter:description" content=""/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/gan/wasserstein/readme.html"/>
|
||||
<meta property="og:title" content="Wasserstein GAN-WGAN"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="Wasserstein GAN-WGAN"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="Wasserstein GAN-WGAN"/>
|
||||
<meta property="og:description" content=""/>
|
||||
|
||||
<title>Wasserstein GAN-WGAN</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/gan/wasserstein/readme.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="../index.html">gan</a>
|
||||
<a class="parent" href="index.html">wasserstein</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/gan/wasserstein/readme.md" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1><a href="https://nn.labml.ai/gan/wasserstein/index.html">Wasserstein GAN-WGAN</a></h1>
|
||||
<p>这是 <a href="https://arxiv.org/abs/1701.07875">Wasserstein GAN</a> 的实现。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
File diff suppressed because it is too large
Load Diff
File diff suppressed because one or more lines are too long
@@ -0,0 +1,130 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content=""/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="图关注网络 (GAT)"/>
|
||||
<meta name="twitter:description" content=""/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/graphs/gat/readme.html"/>
|
||||
<meta property="og:title" content="图关注网络 (GAT)"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="图关注网络 (GAT)"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="图关注网络 (GAT)"/>
|
||||
<meta property="og:description" content=""/>
|
||||
|
||||
<title>图关注网络 (GAT)</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/graphs/gat/readme.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="../index.html">graphs</a>
|
||||
<a class="parent" href="index.html">gat</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/graphs/gat/readme.md" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1><a href="https://nn.labml.ai/graphs/gat/index.html">图表注意力网络 (GAT)</a></h1>
|
||||
<p>这是 <a href="https://pytorch.org">PyTorch</a> 对《<a href="https://arxiv.org/abs/1710.10903">图形注意力网络</a>》论文的实现。</p>
|
||||
<p>GAT 处理图表数据。图由节点和连接节点的边组成。例如，在 Cora 数据集中，节点是研究论文，边缘是连接论文的引文。</p>
|
||||
<p>GAT 使用蒙面自注意力，有点类似于<a href="https://nn.labml.ai/transformers/mha.html">变形金刚</a>。GAT 由相互堆叠的图表注意力层组成。每个图注意力层都将节点嵌入作为转换后的嵌入的输入和输出获得节点。节点嵌入会注意它所连接的其他节点的嵌入。图形注意力层的详细信息与实现一起包括在内。</p>
|
||||
<p>以下是<a href="https://nn.labml.ai/graphs/gat/experiment.html">在 Cora 数据集上训练两层 GAT 的训练代码</a>。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
@@ -0,0 +1,471 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content="这列火车是 Cora 数据集上的 Graph 注意力网络 v2 (GATv2)"/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="在 Cora 数据集上训练图形注意力网络 v2 (GATv2)"/>
|
||||
<meta name="twitter:description" content="这列火车是 Cora 数据集上的 Graph 注意力网络 v2 (GATv2)"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/graphs/gatv2/experiment.html"/>
|
||||
<meta property="og:title" content="在 Cora 数据集上训练图形注意力网络 v2 (GATv2)"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="在 Cora 数据集上训练图形注意力网络 v2 (GATv2)"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="在 Cora 数据集上训练图形注意力网络 v2 (GATv2)"/>
|
||||
<meta property="og:description" content="这列火车是 Cora 数据集上的 Graph 注意力网络 v2 (GATv2)"/>
|
||||
|
||||
<title>在 Cora 数据集上训练图形注意力网络 v2 (GATv2)</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/graphs/gatv2/experiment.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="../index.html">graphs</a>
|
||||
<a class="parent" href="index.html">gatv2</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/graphs/gatv2/experiment.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1>在 Cora 数据集上训练图注意力网络 v2 (Gatv2)</h1>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">11</span><span></span><span class="kn">import</span> <span class="nn">torch</span>
|
||||
<span class="lineno">12</span><span class="kn">from</span> <span class="nn">torch</span> <span class="kn">import</span> <span class="n">nn</span>
|
||||
<span class="lineno">13</span>
|
||||
<span class="lineno">14</span><span class="kn">from</span> <span class="nn">labml</span> <span class="kn">import</span> <span class="n">experiment</span>
|
||||
<span class="lineno">15</span><span class="kn">from</span> <span class="nn">labml.configs</span> <span class="kn">import</span> <span class="n">option</span>
|
||||
<span class="lineno">16</span><span class="kn">from</span> <span class="nn">labml_helpers.module</span> <span class="kn">import</span> <span class="n">Module</span>
|
||||
<span class="lineno">17</span><span class="kn">from</span> <span class="nn">labml_nn.graphs.gat.experiment</span> <span class="kn">import</span> <span class="n">Configs</span> <span class="k">as</span> <span class="n">GATConfigs</span>
|
||||
<span class="lineno">18</span><span class="kn">from</span> <span class="nn">labml_nn.graphs.gatv2</span> <span class="kn">import</span> <span class="n">GraphAttentionV2Layer</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-1'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-1'>#</a>
|
||||
</div>
|
||||
<h2>Graph 注意力网络 v2 (GATv2)</h2>
|
||||
<p>这个图形关注网络有两个<a href="index.html">图形关注层</a>。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">21</span><span class="k">class</span> <span class="nc">GATv2</span><span class="p">(</span><span class="n">Module</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-2'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-2'>#</a>
|
||||
</div>
|
||||
<ul><li><code class="highlight"><span></span><span class="n">in_features</span></code>
|
||||
是每个节点的要素数</li>
|
||||
<li><code class="highlight"><span></span><span class="n">n_hidden</span></code>
|
||||
是第一个图形关注层中的要素数</li>
|
||||
<li><code class="highlight"><span></span><span class="n">n_classes</span></code>
|
||||
是类的数量</li>
|
||||
<li><code class="highlight"><span></span><span class="n">n_heads</span></code>
|
||||
是图表关注层中的头部数量</li>
|
||||
<li><code class="highlight"><span></span><span class="n">dropout</span></code>
|
||||
是辍学概率</li>
|
||||
<li><code class="highlight"><span></span><span class="n">share_weights</span></code>
|
||||
如果设置为 True，则同一矩阵将应用于每条边的源节点和目标节点</li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">28</span> <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">in_features</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">n_hidden</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">n_classes</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">n_heads</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">dropout</span><span class="p">:</span> <span class="nb">float</span><span class="p">,</span>
|
||||
<span class="lineno">29</span> <span class="n">share_weights</span><span class="p">:</span> <span class="nb">bool</span> <span class="o">=</span> <span class="kc">True</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-3'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-3'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">38</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-4'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-4'>#</a>
|
||||
</div>
|
||||
<p>我们连接头部的第一个图形注意层</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">41</span> <span class="bp">self</span><span class="o">.</span><span class="n">layer1</span> <span class="o">=</span> <span class="n">GraphAttentionV2Layer</span><span class="p">(</span><span class="n">in_features</span><span class="p">,</span> <span class="n">n_hidden</span><span class="p">,</span> <span class="n">n_heads</span><span class="p">,</span>
|
||||
<span class="lineno">42</span> <span class="n">is_concat</span><span class="o">=</span><span class="kc">True</span><span class="p">,</span> <span class="n">dropout</span><span class="o">=</span><span class="n">dropout</span><span class="p">,</span> <span class="n">share_weights</span><span class="o">=</span><span class="n">share_weights</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-5'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-5'>#</a>
|
||||
</div>
|
||||
<p>第一个图形关注层之后的激活功能</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">44</span> <span class="bp">self</span><span class="o">.</span><span class="n">activation</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">ELU</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-6'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-6'>#</a>
|
||||
</div>
|
||||
<p>最后一张图关注层，我们平均头部</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">46</span> <span class="bp">self</span><span class="o">.</span><span class="n">output</span> <span class="o">=</span> <span class="n">GraphAttentionV2Layer</span><span class="p">(</span><span class="n">n_hidden</span><span class="p">,</span> <span class="n">n_classes</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span>
|
||||
<span class="lineno">47</span> <span class="n">is_concat</span><span class="o">=</span><span class="kc">False</span><span class="p">,</span> <span class="n">dropout</span><span class="o">=</span><span class="n">dropout</span><span class="p">,</span> <span class="n">share_weights</span><span class="o">=</span><span class="n">share_weights</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-7'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-7'>#</a>
|
||||
</div>
|
||||
<p>辍学</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">49</span> <span class="bp">self</span><span class="o">.</span><span class="n">dropout</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Dropout</span><span class="p">(</span><span class="n">dropout</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-8'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-8'>#</a>
|
||||
</div>
|
||||
<ul><li><code class="highlight"><span></span><span class="n">x</span></code>
|
||||
是形状的特征向量<code class="highlight"><span></span><span class="p">[</span><span class="n">n_nodes</span><span class="p">,</span> <span class="n">in_features</span><span class="p">]</span></code>
|
||||
</li>
|
||||
<li><code class="highlight"><span></span><span class="n">adj_mat</span></code>
|
||||
是形式的邻接矩阵<code class="highlight"><span></span><span class="p">[</span><span class="n">n_nodes</span><span class="p">,</span> <span class="n">n_nodes</span><span class="p">,</span> <span class="n">n_heads</span><span class="p">]</span></code>
|
||||
或<code class="highlight"><span></span><span class="p">[</span><span class="n">n_nodes</span><span class="p">,</span> <span class="n">n_nodes</span><span class="p">,</span> <span class="mi">1</span><span class="p">]</span></code>
|
||||
</li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">51</span> <span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span> <span class="n">adj_mat</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-9'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-9'>#</a>
|
||||
</div>
|
||||
<p>将丢失应用于输入</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">58</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">dropout</span><span class="p">(</span><span class="n">x</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-10'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-10'>#</a>
|
||||
</div>
|
||||
<p>第一个图形关注层</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">60</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">layer1</span><span class="p">(</span><span class="n">x</span><span class="p">,</span> <span class="n">adj_mat</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-11'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-11'>#</a>
|
||||
</div>
|
||||
<p>激活功能</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">62</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">activation</span><span class="p">(</span><span class="n">x</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-12'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-12'>#</a>
|
||||
</div>
|
||||
<p>辍学</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">64</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">dropout</span><span class="p">(</span><span class="n">x</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-13'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-13'>#</a>
|
||||
</div>
|
||||
<p>logits 的输出层（未激活）</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">66</span> <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">output</span><span class="p">(</span><span class="n">x</span><span class="p">,</span> <span class="n">adj_mat</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-14'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-14'>#</a>
|
||||
</div>
|
||||
<h2>配置</h2>
|
||||
<p>由于实验与 <a href="../gat/experiment.html">GAT 实验</a>相同，但使用 G <a href="index.html">ATv2 模型</a>，我们扩展了相同的配置并更改了模型。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">69</span><span class="k">class</span> <span class="nc">Configs</span><span class="p">(</span><span class="n">GATConfigs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-15'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-15'>#</a>
|
||||
</div>
|
||||
<p>是否共享边的源节点和目标节点的权重</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">78</span> <span class="n">share_weights</span><span class="p">:</span> <span class="nb">bool</span> <span class="o">=</span> <span class="kc">False</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-16'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-16'>#</a>
|
||||
</div>
|
||||
<p>设置模型</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">80</span> <span class="n">model</span><span class="p">:</span> <span class="n">GATv2</span> <span class="o">=</span> <span class="s1">'gat_v2_model'</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-17'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-17'>#</a>
|
||||
</div>
|
||||
<p>创建 GATv2 模型</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">83</span><span class="nd">@option</span><span class="p">(</span><span class="n">Configs</span><span class="o">.</span><span class="n">model</span><span class="p">)</span>
|
||||
<span class="lineno">84</span><span class="k">def</span> <span class="nf">gat_v2_model</span><span class="p">(</span><span class="n">c</span><span class="p">:</span> <span class="n">Configs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-18'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-18'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">88</span> <span class="k">return</span> <span class="n">GATv2</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">in_features</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">n_hidden</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">n_classes</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">n_heads</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">dropout</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">share_weights</span><span class="p">)</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">device</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-19'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-19'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">91</span><span class="k">def</span> <span class="nf">main</span><span class="p">():</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-20'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-20'>#</a>
|
||||
</div>
|
||||
<p>创建配置</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">93</span> <span class="n">conf</span> <span class="o">=</span> <span class="n">Configs</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-21'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-21'>#</a>
|
||||
</div>
|
||||
<p>创建实验</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">95</span> <span class="n">experiment</span><span class="o">.</span><span class="n">create</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">'gatv2'</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-22'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-22'>#</a>
|
||||
</div>
|
||||
<p>计算配置。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">97</span> <span class="n">experiment</span><span class="o">.</span><span class="n">configs</span><span class="p">(</span><span class="n">conf</span><span class="p">,</span> <span class="p">{</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-23'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-23'>#</a>
|
||||
</div>
|
||||
<p>Adam 优化器</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">99</span> <span class="s1">'optimizer.optimizer'</span><span class="p">:</span> <span class="s1">'Adam'</span><span class="p">,</span>
|
||||
<span class="lineno">100</span> <span class="s1">'optimizer.learning_rate'</span><span class="p">:</span> <span class="mf">5e-3</span><span class="p">,</span>
|
||||
<span class="lineno">101</span> <span class="s1">'optimizer.weight_decay'</span><span class="p">:</span> <span class="mf">5e-4</span><span class="p">,</span>
|
||||
<span class="lineno">102</span>
|
||||
<span class="lineno">103</span> <span class="s1">'dropout'</span><span class="p">:</span> <span class="mf">0.7</span><span class="p">,</span>
|
||||
<span class="lineno">104</span> <span class="p">})</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-24'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-24'>#</a>
|
||||
</div>
|
||||
<p>开始观看实验</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">107</span> <span class="k">with</span> <span class="n">experiment</span><span class="o">.</span><span class="n">start</span><span class="p">():</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-25'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-25'>#</a>
|
||||
</div>
|
||||
<p>运行训练</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">109</span> <span class="n">conf</span><span class="o">.</span><span class="n">run</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-26'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-26'>#</a>
|
||||
</div>
|
||||
<p></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">113</span><span class="k">if</span> <span class="vm">__name__</span> <span class="o">==</span> <span class="s1">'__main__'</span><span class="p">:</span>
|
||||
<span class="lineno">114</span> <span class="n">main</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
File diff suppressed because one or more lines are too long
@@ -0,0 +1,130 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content=""/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="Graph 注意力网络 v2 (GATv2)"/>
|
||||
<meta name="twitter:description" content=""/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/graphs/gatv2/readme.html"/>
|
||||
<meta property="og:title" content="Graph 注意力网络 v2 (GATv2)"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="Graph 注意力网络 v2 (GATv2)"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="Graph 注意力网络 v2 (GATv2)"/>
|
||||
<meta property="og:description" content=""/>
|
||||
|
||||
<title>Graph 注意力网络 v2 (GATv2)</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/graphs/gatv2/readme.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="../index.html">graphs</a>
|
||||
<a class="parent" href="index.html">gatv2</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/graphs/gatv2/readme.md" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1><a href="https://nn.labml.ai/graphs/gatv2/index.html">Graph 注意力网络 v2 (Gatv2)</a></h1>
|
||||
<p>这是 <a href="https://pytorch.org">PyTorch</a> 对 Gatv2 运算符的实现，摘自《<a href="https://arxiv.org/abs/2105.14491">图注意力网络有多专心？</a>》</p>。
|
||||
<p>Gatv2 处理图表数据。图由节点和连接节点的边组成。例如，在 Cora 数据集中，节点是研究论文，边缘是连接论文的引文。</p>
|
||||
G@@ <p>atv2 运算符修复了标准 GAT 的静态注意力问题：由于标准 GAT 中的线性层是紧接应用的，因此有人值守节点的排名不受查询节点的限制。相比之下，在 Gatv2 中，每个节点都可以管理任何其他节点。</p>
|
||||
<p>以下是<a href="https://nn.labml.ai/graphs/gatv2/experiment.html">在 Cora 数据集上训练双层 Gatv2 的训练代码</a>。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
@@ -0,0 +1,127 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content="一组与图神经网络相关的 PyTorch 实现/教程"/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="绘制神经网络"/>
|
||||
<meta name="twitter:description" content="一组与图神经网络相关的 PyTorch 实现/教程"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/graphs/index.html"/>
|
||||
<meta property="og:title" content="绘制神经网络"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="绘制神经网络"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="绘制神经网络"/>
|
||||
<meta property="og:description" content="一组与图神经网络相关的 PyTorch 实现/教程"/>
|
||||
|
||||
<title>绘制神经网络</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/graphs/index.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="index.html">graphs</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/graphs/__init__.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1>绘制神经网络</h1>
|
||||
<ul><li><a href="gat/index.html">图关注网络 (GAT)</a></li>
|
||||
<li><a href="gatv2/index.html">Graph 注意力网络 v2 (GATv2)</a></li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
@@ -0,0 +1,387 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="zh">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content=""/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="experiment.py"/>
|
||||
<meta name="twitter:description" content=""/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/hypernetworks/experiment.html"/>
|
||||
<meta property="og:title" content="experiment.py"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="experiment.py"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="experiment.py"/>
|
||||
<meta property="og:description" content=""/>
|
||||
|
||||
<title>experiment.py</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/hypernetworks/experiment.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="index.html">hypernetworks</a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations" target="_blank">
|
||||
<img alt="Github"
|
||||
src="https://img.shields.io/github/stars/labmlai/annotated_deep_learning_paper_implementations?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
<a href="https://twitter.com/labmlai" rel="nofollow" target="_blank">
|
||||
<img alt="Twitter"
|
||||
src="https://img.shields.io/twitter/follow/labmlai?style=social"
|
||||
style="max-width:100%;"/></a>
|
||||
</p>
|
||||
<p>
|
||||
<a href="https://github.com/labmlai/annotated_deep_learning_paper_implementations/tree/master/labml_nn/hypernetworks/experiment.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">1</span><span></span><span class="kn">import</span> <span class="nn">torch</span>
|
||||
<span class="lineno">2</span><span class="kn">import</span> <span class="nn">torch.nn</span> <span class="k">as</span> <span class="nn">nn</span>
|
||||
<span class="lineno">3</span><span class="kn">from</span> <span class="nn">labml</span> <span class="kn">import</span> <span class="n">experiment</span>
|
||||
<span class="lineno">4</span><span class="kn">from</span> <span class="nn">labml.configs</span> <span class="kn">import</span> <span class="n">option</span>
|
||||
<span class="lineno">5</span><span class="kn">from</span> <span class="nn">labml.utils.pytorch</span> <span class="kn">import</span> <span class="n">get_modules</span>
|
||||
<span class="lineno">6</span><span class="kn">from</span> <span class="nn">labml_helpers.module</span> <span class="kn">import</span> <span class="n">Module</span>
|
||||
<span class="lineno">7</span>
|
||||
<span class="lineno">8</span><span class="kn">from</span> <span class="nn">labml_nn.experiments.nlp_autoregression</span> <span class="kn">import</span> <span class="n">NLPAutoRegressionConfigs</span>
|
||||
<span class="lineno">9</span><span class="kn">from</span> <span class="nn">labml_nn.hypernetworks.hyper_lstm</span> <span class="kn">import</span> <span class="n">HyperLSTM</span>
|
||||
<span class="lineno">10</span><span class="kn">from</span> <span class="nn">labml_nn.lstm</span> <span class="kn">import</span> <span class="n">LSTM</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-1'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-1'>#</a>
|
||||
</div>
|
||||
<h2>自动回归模型</h2>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">13</span><span class="k">class</span> <span class="nc">AutoregressiveModel</span><span class="p">(</span><span class="n">Module</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-2'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-2'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">18</span> <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">n_vocab</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">d_model</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">rnn_model</span><span class="p">:</span> <span class="n">Module</span><span class="p">):</span>
|
||||
<span class="lineno">19</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-3'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-3'>#</a>
|
||||
</div>
|
||||
<p>令牌嵌入模块</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">21</span> <span class="bp">self</span><span class="o">.</span><span class="n">src_embed</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Embedding</span><span class="p">(</span><span class="n">n_vocab</span><span class="p">,</span> <span class="n">d_model</span><span class="p">)</span>
|
||||
<span class="lineno">22</span> <span class="bp">self</span><span class="o">.</span><span class="n">lstm</span> <span class="o">=</span> <span class="n">rnn_model</span>
|
||||
<span class="lineno">23</span> <span class="bp">self</span><span class="o">.</span><span class="n">generator</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Linear</span><span class="p">(</span><span class="n">d_model</span><span class="p">,</span> <span class="n">n_vocab</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-4'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-4'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">25</span> <span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">):</span>
|
||||
<span class="lineno">26</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">src_embed</span><span class="p">(</span><span class="n">x</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-5'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-5'>#</a>
|
||||
</div>
|
||||
<p>嵌入令牌 (<code class="highlight"><span></span><span class="n">src</span></code>
|
||||
) 并通过变压器运行它</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">28</span> <span class="n">res</span><span class="p">,</span> <span class="n">state</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">lstm</span><span class="p">(</span><span class="n">x</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-6'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-6'>#</a>
|
||||
</div>
|
||||
<p>生成下一个令牌的日志</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">30</span> <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">generator</span><span class="p">(</span><span class="n">res</span><span class="p">),</span> <span class="n">state</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-7'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-7'>#</a>
|
||||
</div>
|
||||
<h2>配置</h2>
|
||||
<p>当我们开始实验时，默认配置可以而且将会被覆盖</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">33</span><span class="k">class</span> <span class="nc">Configs</span><span class="p">(</span><span class="n">NLPAutoRegressionConfigs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-8'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-8'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">40</span> <span class="n">model</span><span class="p">:</span> <span class="n">AutoregressiveModel</span>
|
||||
<span class="lineno">41</span> <span class="n">rnn_model</span><span class="p">:</span> <span class="n">Module</span>
|
||||
<span class="lineno">42</span>
|
||||
<span class="lineno">43</span> <span class="n">d_model</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">512</span>
|
||||
<span class="lineno">44</span> <span class="n">n_rhn</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">16</span>
|
||||
<span class="lineno">45</span> <span class="n">n_z</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">16</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-9'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-9'>#</a>
|
||||
</div>
|
||||
<p>初始化自回归模型</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">48</span><span class="nd">@option</span><span class="p">(</span><span class="n">Configs</span><span class="o">.</span><span class="n">model</span><span class="p">)</span>
|
||||
<span class="lineno">49</span><span class="k">def</span> <span class="nf">autoregressive_model</span><span class="p">(</span><span class="n">c</span><span class="p">:</span> <span class="n">Configs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-10'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-10'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">53</span> <span class="n">m</span> <span class="o">=</span> <span class="n">AutoregressiveModel</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">n_tokens</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">d_model</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">rnn_model</span><span class="p">)</span>
|
||||
<span class="lineno">54</span> <span class="k">return</span> <span class="n">m</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">device</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-11'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-11'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">57</span><span class="nd">@option</span><span class="p">(</span><span class="n">Configs</span><span class="o">.</span><span class="n">rnn_model</span><span class="p">)</span>
|
||||
<span class="lineno">58</span><span class="k">def</span> <span class="nf">hyper_lstm</span><span class="p">(</span><span class="n">c</span><span class="p">:</span> <span class="n">Configs</span><span class="p">):</span>
|
||||
<span class="lineno">59</span> <span class="k">return</span> <span class="n">HyperLSTM</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">d_model</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">d_model</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">n_rhn</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">n_z</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span>
|
||||
<span class="lineno">60</span>
|
||||
<span class="lineno">61</span>
|
||||
<span class="lineno">62</span><span class="nd">@option</span><span class="p">(</span><span class="n">Configs</span><span class="o">.</span><span class="n">rnn_model</span><span class="p">)</span>
|
||||
<span class="lineno">63</span><span class="k">def</span> <span class="nf">lstm</span><span class="p">(</span><span class="n">c</span><span class="p">:</span> <span class="n">Configs</span><span class="p">):</span>
|
||||
<span class="lineno">64</span> <span class="k">return</span> <span class="n">LSTM</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">d_model</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">d_model</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span>
|
||||
<span class="lineno">65</span>
|
||||
<span class="lineno">66</span>
|
||||
<span class="lineno">67</span><span class="k">def</span> <span class="nf">main</span><span class="p">():</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-12'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-12'>#</a>
|
||||
</div>
|
||||
<p>创建实验</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">69</span> <span class="n">experiment</span><span class="o">.</span><span class="n">create</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s2">"hyper_lstm"</span><span class="p">,</span> <span class="n">comment</span><span class="o">=</span><span class="s1">''</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-13'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-13'>#</a>
|
||||
</div>
|
||||
<p>创建配置</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">71</span> <span class="n">conf</span> <span class="o">=</span> <span class="n">Configs</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-14'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-14'>#</a>
|
||||
</div>
|
||||
<p>装载配置</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">73</span> <span class="n">experiment</span><span class="o">.</span><span class="n">configs</span><span class="p">(</span><span class="n">conf</span><span class="p">,</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-15'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-15'>#</a>
|
||||
</div>
|
||||
<p>要覆盖的配置字典</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">75</span> <span class="p">{</span><span class="s1">'tokenizer'</span><span class="p">:</span> <span class="s1">'character'</span><span class="p">,</span>
|
||||
<span class="lineno">76</span> <span class="s1">'text'</span><span class="p">:</span> <span class="s1">'tiny_shakespeare'</span><span class="p">,</span>
|
||||
<span class="lineno">77</span> <span class="s1">'optimizer.learning_rate'</span><span class="p">:</span> <span class="mf">2.5e-4</span><span class="p">,</span>
|
||||
<span class="lineno">78</span> <span class="s1">'optimizer.optimizer'</span><span class="p">:</span> <span class="s1">'Adam'</span><span class="p">,</span>
|
||||
<span class="lineno">79</span> <span class="s1">'prompt'</span><span class="p">:</span> <span class="s1">'It is'</span><span class="p">,</span>
|
||||
<span class="lineno">80</span> <span class="s1">'prompt_separator'</span><span class="p">:</span> <span class="s1">''</span><span class="p">,</span>
|
||||
<span class="lineno">81</span>
|
||||
<span class="lineno">82</span> <span class="s1">'rnn_model'</span><span class="p">:</span> <span class="s1">'hyper_lstm'</span><span class="p">,</span>
|
||||
<span class="lineno">83</span>
|
||||
<span class="lineno">84</span> <span class="s1">'train_loader'</span><span class="p">:</span> <span class="s1">'shuffled_train_loader'</span><span class="p">,</span>
|
||||
<span class="lineno">85</span> <span class="s1">'valid_loader'</span><span class="p">:</span> <span class="s1">'shuffled_valid_loader'</span><span class="p">,</span>
|
||||
<span class="lineno">86</span>
|
||||
<span class="lineno">87</span> <span class="s1">'seq_len'</span><span class="p">:</span> <span class="mi">512</span><span class="p">,</span>
|
||||
<span class="lineno">88</span> <span class="s1">'epochs'</span><span class="p">:</span> <span class="mi">128</span><span class="p">,</span>
|
||||
<span class="lineno">89</span> <span class="s1">'batch_size'</span><span class="p">:</span> <span class="mi">2</span><span class="p">,</span>
|
||||
<span class="lineno">90</span> <span class="s1">'inner_iterations'</span><span class="p">:</span> <span class="mi">25</span><span class="p">})</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-16'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-16'>#</a>
|
||||
</div>
|
||||
<p>设置用于保存和加载的模型</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">93</span> <span class="n">experiment</span><span class="o">.</span><span class="n">add_pytorch_models</span><span class="p">(</span><span class="n">get_modules</span><span class="p">(</span><span class="n">conf</span><span class="p">))</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-17'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-17'>#</a>
|
||||
</div>
|
||||
<p>开始实验</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">96</span> <span class="k">with</span> <span class="n">experiment</span><span class="o">.</span><span class="n">start</span><span class="p">():</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-18'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-18'>#</a>
|
||||
</div>
|
||||
<p><code class="highlight"><span></span><span class="n">TrainValidConfigs</span><span class="o">.</span><span class="n">run</span></code>
|
||||
</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">98</span> <span class="n">conf</span><span class="o">.</span><span class="n">run</span><span class="p">()</span>
|
||||
<span class="lineno">99</span>
|
||||
<span class="lineno">100</span>
|
||||
<span class="lineno">101</span><span class="k">if</span> <span class="vm">__name__</span> <span class="o">==</span> <span class="s1">'__main__'</span><span class="p">:</span>
|
||||
<span class="lineno">102</span> <span class="n">main</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
Some files were not shown because too many files have changed in this diff Show More
Reference in New Issue
Block a user