chore: import upstream snapshot with attribution
This commit is contained in:
wehub-resource-sync
@@ -0,0 +1,521 @@
|
||||
<!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="使用冻结 BERT 模型为区块生成 BERT 嵌入"/>
|
||||
|
||||
<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="BERT 文本块的嵌入"/>
|
||||
<meta name="twitter:description" content="使用冻结 BERT 模型为区块生成 BERT 嵌入"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/transformers/retro/bert_embeddings.html"/>
|
||||
<meta property="og:title" content="BERT 文本块的嵌入"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="BERT 文本块的嵌入"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="BERT 文本块的嵌入"/>
|
||||
<meta property="og:description" content="使用冻结 BERT 模型为区块生成 BERT 嵌入"/>
|
||||
|
||||
<title>BERT 文本块的嵌入</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/transformers/retro/bert_embeddings.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">transformers</a>
|
||||
<a class="parent" href="index.html">retro</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/transformers/retro/bert_embeddings.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>BERT 文本块的嵌入</h1>
|
||||
<p>这是获取 <a href="index.html">RETRO 模型</a>块的 BERT 嵌入的代码。</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">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">transformers</span> <span class="kn">import</span> <span class="n">BertTokenizer</span><span class="p">,</span> <span class="n">BertModel</span>
|
||||
<span class="lineno">17</span>
|
||||
<span class="lineno">18</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">monit</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>BERT 嵌入式</h2>
|
||||
<p>对于给定的文本块，<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 eqb" style=""><span class="mord mathnormal" style="margin-right:0.10903em">N</span></span></span></span></span></span>这个类会生成 BERT 嵌入<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 eqa" style=""><span class="mord text" style=""><span class="mord" style="">B</span><span class="mord sizing reset-size6 size5" style=""><span class="mord" style="">ERT</span></span></span><span class="mopen" style="">(</span><span class="mord" style=""><span class="mord mathnormal coloredeq eqb" style="margin-right:0.10903em">N</span></span><span class="mclose" style="">)</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 eqa" style=""><span class="mord text" style=""><span class="mord" style="">B</span><span class="mord sizing reset-size6 size5" style=""><span class="mord" style="">ERT</span></span></span><span class="mopen" style="">(</span><span class="mord" style=""><span class="mord mathnormal coloredeq eqb" style="margin-right:0.10903em">N</span></span><span class="mclose" style="">)</span></span></span></span></span></span>是所有令牌的 BERT 嵌入的平均值<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 eqb" style=""><span class="mord mathnormal" style="margin-right:0.10903em">N</span></span></span></span></span></span>。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">21</span><span class="k">class</span> <span class="nc">BERTChunkEmbeddings</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">29</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">device</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">device</span><span class="p">):</span>
|
||||
<span class="lineno">30</span> <span class="bp">self</span><span class="o">.</span><span class="n">device</span> <span class="o">=</span> <span class="n">device</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>从 <a href="https://huggingface.co/bert-base-uncased">HuggingFac</a> e 加载 BERT 分词器</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">33</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 BERT tokenizer'</span><span class="p">):</span>
|
||||
<span class="lineno">34</span> <span class="bp">self</span><span class="o">.</span><span class="n">tokenizer</span> <span class="o">=</span> <span class="n">BertTokenizer</span><span class="o">.</span><span class="n">from_pretrained</span><span class="p">(</span><span class="s1">'bert-base-uncased'</span><span class="p">,</span>
|
||||
<span class="lineno">35</span> <span class="n">cache_dir</span><span class="o">=</span><span class="nb">str</span><span class="p">(</span>
|
||||
<span class="lineno">36</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">'cache'</span> <span class="o">/</span> <span class="s1">'bert-tokenizer'</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>从 <a href="https://huggingface.co/bert-base-uncased">HuggingFac</a> e 加载 BERT 模型</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">39</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 BERT model'</span><span class="p">):</span>
|
||||
<span class="lineno">40</span> <span class="bp">self</span><span class="o">.</span><span class="n">model</span> <span class="o">=</span> <span class="n">BertModel</span><span class="o">.</span><span class="n">from_pretrained</span><span class="p">(</span><span class="s2">"bert-base-uncased"</span><span class="p">,</span>
|
||||
<span class="lineno">41</span> <span class="n">cache_dir</span><span class="o">=</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="o">/</span> <span class="s1">'cache'</span> <span class="o">/</span> <span class="s1">'bert-model'</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>将模型移到<code class="highlight"><span></span><span class="n">device</span></code>
|
||||
</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">model</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">device</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>在此实现中，我们不会使用固定数量的令牌制作区块。原因之一是此实现使用字符级令牌，而 BERT 使用其子词分词器。</p>
|
||||
<p>因此，此方法将截断文本以确保没有部分标记。</p>
|
||||
<p>例如，一个块可能像<code class="highlight"><span></span><span class="n">s</span> <span class="n">a</span> <span class="n">popular</span> <span class="n">programming</span> <span class="n">la</span></code>
|
||||
，末尾带有部分单词（部分子词标记）。我们剥离它们以获得更好的 BERT 嵌入。如前所述，如果我们在标记化后破坏了区块，则没有必要这样做。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">46</span> <span class="nd">@staticmethod</span>
|
||||
<span class="lineno">47</span> <span class="k">def</span> <span class="nf">_trim_chunk</span><span class="p">(</span><span class="n">chunk</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>去掉空白</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">61</span> <span class="n">stripped</span> <span class="o">=</span> <span class="n">chunk</span><span class="o">.</span><span class="n">strip</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>断词</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">63</span> <span class="n">parts</span> <span class="o">=</span> <span class="n">stripped</span><span class="o">.</span><span class="n">split</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>移除第一块和最后一块碎片</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">65</span> <span class="n">stripped</span> <span class="o">=</span> <span class="n">stripped</span><span class="p">[</span><span class="nb">len</span><span class="p">(</span><span class="n">parts</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">parts</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-10'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-10'>#</a>
|
||||
</div>
|
||||
<p>移除空格</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">68</span> <span class="n">stripped</span> <span class="o">=</span> <span class="n">stripped</span><span class="o">.</span><span class="n">strip</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>如果为空则返回原始字符串</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">71</span> <span class="k">if</span> <span class="ow">not</span> <span class="n">stripped</span><span class="p">:</span>
|
||||
<span class="lineno">72</span> <span class="k">return</span> <span class="n">chunk</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>否则，返回被剥离的字符串</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">74</span> <span class="k">else</span><span class="p">:</span>
|
||||
<span class="lineno">75</span> <span class="k">return</span> <span class="n">stripped</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><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 eqa" style=""><span class="mord text" style=""><span class="mord" style="">B</span><span class="mord sizing reset-size6 size5" style=""><span class="mord" style="">ERT</span></span></span><span class="mopen" style="">(</span><span class="mord" style=""><span class="mord mathnormal coloredeq eqb" style="margin-right:0.10903em">N</span></span><span class="mclose" style="">)</span></span></span></span></span></span>获取区块列表。</h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">77</span> <span class="k">def</span> <span class="fm">__call__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">chunks</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-14'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-14'>#</a>
|
||||
</div>
|
||||
<p>我们不需要计算梯度</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">83</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-15'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-15'>#</a>
|
||||
</div>
|
||||
<p>修剪块</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">85</span> <span class="n">trimmed_chunks</span> <span class="o">=</span> <span class="p">[</span><span class="bp">self</span><span class="o">.</span><span class="n">_trim_chunk</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">chunks</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>使用 BERT 分词器对区块进行标记化</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">88</span> <span class="n">tokens</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">trimmed_chunks</span><span class="p">,</span> <span class="n">return_tensors</span><span class="o">=</span><span class="s1">'pt'</span><span class="p">,</span> <span class="n">add_special_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="kc">True</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>将令牌 ID、注意掩码和令牌类型移动到设备</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">91</span> <span class="n">input_ids</span> <span class="o">=</span> <span class="n">tokens</span><span class="p">[</span><span class="s1">'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>
|
||||
<span class="lineno">92</span> <span class="n">attention_mask</span> <span class="o">=</span> <span class="n">tokens</span><span class="p">[</span><span class="s1">'attention_mask'</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="lineno">93</span> <span class="n">token_type_ids</span> <span class="o">=</span> <span class="n">tokens</span><span class="p">[</span><span class="s1">'token_type_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-18'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-18'>#</a>
|
||||
</div>
|
||||
<p>评估模型</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">95</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">input_ids</span><span class="o">=</span><span class="n">input_ids</span><span class="p">,</span>
|
||||
<span class="lineno">96</span> <span class="n">attention_mask</span><span class="o">=</span><span class="n">attention_mask</span><span class="p">,</span>
|
||||
<span class="lineno">97</span> <span class="n">token_type_ids</span><span class="o">=</span><span class="n">token_type_ids</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>获取令牌嵌入</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">100</span> <span class="n">state</span> <span class="o">=</span> <span class="n">output</span><span class="p">[</span><span class="s1">'last_hidden_state'</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>计算平均代币嵌入量。请注意，<code class="highlight"><span></span><span class="mi">0</span></code>
|
||||
如果令牌是空填充的，则注意掩码为。我们得到空令牌，因为这些块的长度不同。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">104</span> <span class="n">emb</span> <span class="o">=</span> <span class="p">(</span><span class="n">state</span> <span class="o">*</span> <span class="n">attention_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">sum</span><span class="p">(</span><span class="n">dim</span><span class="o">=</span><span class="mi">1</span><span class="p">)</span> <span class="o">/</span> <span class="n">attention_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">sum</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-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">107</span> <span class="k">return</span> <span class="n">emb</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>
|
||||
<h3>用于测试 BERT 嵌入的代码</h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">110</span><span class="k">def</span> <span class="nf">_test</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">114</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-24'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-24'>#</a>
|
||||
</div>
|
||||
<p>初始化</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">117</span> <span class="n">device</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">device</span><span class="p">(</span><span class="s1">'cuda:0'</span><span class="p">)</span>
|
||||
<span class="lineno">118</span> <span class="n">bert</span> <span class="o">=</span> <span class="n">BERTChunkEmbeddings</span><span class="p">(</span><span class="n">device</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>样本</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">121</span> <span class="n">text</span> <span class="o">=</span> <span class="p">[</span><span class="s2">"Replace me by any text you'd like."</span><span class="p">,</span>
|
||||
<span class="lineno">122</span> <span class="s2">"Second sentence"</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>查看 BERT 分词器</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">125</span> <span class="n">encoded_input</span> <span class="o">=</span> <span class="n">bert</span><span class="o">.</span><span class="n">tokenizer</span><span class="p">(</span><span class="n">text</span><span class="p">,</span> <span class="n">return_tensors</span><span class="o">=</span><span class="s1">'pt'</span><span class="p">,</span> <span class="n">add_special_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="kc">True</span><span class="p">)</span>
|
||||
<span class="lineno">126</span>
|
||||
<span class="lineno">127</span> <span class="n">inspect</span><span class="p">(</span><span class="n">encoded_input</span><span class="p">,</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-27'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-27'>#</a>
|
||||
</div>
|
||||
<p>检查 BERT 模型输出</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">130</span> <span class="n">output</span> <span class="o">=</span> <span class="n">bert</span><span class="o">.</span><span class="n">model</span><span class="p">(</span><span class="n">input_ids</span><span class="o">=</span><span class="n">encoded_input</span><span class="p">[</span><span class="s1">'input_ids'</span><span class="p">]</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">device</span><span class="p">),</span>
|
||||
<span class="lineno">131</span> <span class="n">attention_mask</span><span class="o">=</span><span class="n">encoded_input</span><span class="p">[</span><span class="s1">'attention_mask'</span><span class="p">]</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">device</span><span class="p">),</span>
|
||||
<span class="lineno">132</span> <span class="n">token_type_ids</span><span class="o">=</span><span class="n">encoded_input</span><span class="p">[</span><span class="s1">'token_type_ids'</span><span class="p">]</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">device</span><span class="p">))</span>
|
||||
<span class="lineno">133</span>
|
||||
<span class="lineno">134</span> <span class="n">inspect</span><span class="p">({</span><span class="s1">'last_hidden_state'</span><span class="p">:</span> <span class="n">output</span><span class="p">[</span><span class="s1">'last_hidden_state'</span><span class="p">],</span>
|
||||
<span class="lineno">135</span> <span class="s1">'pooler_output'</span><span class="p">:</span> <span class="n">output</span><span class="p">[</span><span class="s1">'pooler_output'</span><span class="p">]},</span>
|
||||
<span class="lineno">136</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-28'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-28'>#</a>
|
||||
</div>
|
||||
<p>检查从令牌 ID 中重新创建文本</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">139</span> <span class="n">inspect</span><span class="p">(</span><span class="n">bert</span><span class="o">.</span><span class="n">tokenizer</span><span class="o">.</span><span class="n">convert_ids_to_tokens</span><span class="p">(</span><span class="n">encoded_input</span><span class="p">[</span><span class="s1">'input_ids'</span><span class="p">][</span><span class="mi">0</span><span class="p">]),</span> <span class="n">_n</span><span class="o">=-</span><span class="mi">1</span><span class="p">)</span>
|
||||
<span class="lineno">140</span> <span class="n">inspect</span><span class="p">(</span><span class="n">bert</span><span class="o">.</span><span class="n">tokenizer</span><span class="o">.</span><span class="n">convert_ids_to_tokens</span><span class="p">(</span><span class="n">encoded_input</span><span class="p">[</span><span class="s1">'input_ids'</span><span class="p">][</span><span class="mi">1</span><span class="p">]),</span> <span class="n">_n</span><span class="o">=-</span><span class="mi">1</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>获取区块嵌入</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">143</span> <span class="n">inspect</span><span class="p">(</span><span class="n">bert</span><span class="p">(</span><span class="n">text</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></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">147</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">148</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 +1,535 @@
|
||||
<!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="最近邻检索和创建数据库"/>
|
||||
|
||||
<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="最近邻检索的数据库"/>
|
||||
<meta name="twitter:description" content="最近邻检索和创建数据库"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/transformers/retro/database.html"/>
|
||||
<meta property="og:title" content="最近邻检索的数据库"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="最近邻检索的数据库"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="最近邻检索的数据库"/>
|
||||
<meta property="og:description" content="最近邻检索和创建数据库"/>
|
||||
|
||||
<title>最近邻检索的数据库</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/transformers/retro/database.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">transformers</a>
|
||||
<a class="parent" href="index.html">retro</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/transformers/retro/database.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>最近邻检索的数据库</h1>
|
||||
<p>这是构建数据库并检索 RETRO <a href="index.html">模型的最近邻域</a>。</p>
|
||||
</a><p>我们使用 <a href="https://faiss.ai/">FAISS 库作为数据库，而论文使用了 ScanN 库。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">16</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">Optional</span>
|
||||
<span class="lineno">17</span>
|
||||
<span class="lineno">18</span><span class="kn">import</span> <span class="nn">faiss</span>
|
||||
<span class="lineno">19</span><span class="kn">import</span> <span class="nn">numpy</span> <span class="k">as</span> <span class="nn">np</span>
|
||||
<span class="lineno">20</span><span class="kn">import</span> <span class="nn">torch</span>
|
||||
<span class="lineno">21</span>
|
||||
<span class="lineno">22</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">monit</span>
|
||||
<span class="lineno">23</span><span class="kn">from</span> <span class="nn">labml_helpers.datasets.text</span> <span class="kn">import</span> <span class="n">TextFileDataset</span>
|
||||
<span class="lineno">24</span><span class="kn">from</span> <span class="nn">labml_nn.transformers.retro.bert_embeddings</span> <span class="kn">import</span> <span class="n">BERTChunkEmbeddings</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>建立数据库</h2>
|
||||
<ul><li><code class="highlight"><span></span><span class="n">chunk_len</span></code>
|
||||
是块的长度（字符数）</li>
|
||||
<li><code class="highlight"><span></span><span class="n">batch_size</span></code>
|
||||
是计算时要使用的批次大小<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 eqa" style=""><span class="mord text" style=""><span class="mord" style="">B</span><span class="mord sizing reset-size6 size5" style=""><span class="mord" style="">ERT</span></span></span><span class="mopen" style="">(</span><span class="mord mathnormal" style="margin-right:0.10903em">N</span><span class="mclose" style="">)</span></span></span></span></span></span></li>
|
||||
<li><code class="highlight"><span></span><span class="n">d_emb</span></code>
|
||||
是要在 <a href="https://faiss.ai/cpp_api/struct/structfaiss_1_1IndexIVFPQ.html">FAISS 索引中选择的<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 eqa" style=""><span class="mord text" style=""><span class="mord" style="">B</span><span class="mord sizing reset-size6 size5" style=""><span class="mord" style="">ERT</span></span></span><span class="mopen" style="">(</span><span class="mord mathnormal" style="margin-right:0.10903em">N</span><span class="mclose" style="">)</span></span></span></span></span></span>嵌入列表中要</a>素的数量</li>
|
||||
<li><code class="highlight"><span></span><span class="n">n_centeroids</span></code>
|
||||
是索引中的列表数</li>
|
||||
<li><code class="highlight"><span></span><span class="n">code_size</span></code>
|
||||
索引中的编码向量大小</li>
|
||||
<li><code class="highlight"><span></span><span class="n">n_probe</span></code>
|
||||
是要探测的列表的数量</li>
|
||||
<li>`n_train'是用于训练索引的键的数量</li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">27</span><span class="k">def</span> <span class="nf">build_database</span><span class="p">(</span><span class="n">chunk_len</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">16</span><span class="p">,</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><span class="p">,</span> <span class="n">d_emb</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">768</span><span class="p">,</span> <span class="n">n_centeroids</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">256</span><span class="p">,</span>
|
||||
<span class="lineno">28</span> <span class="n">code_size</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">64</span><span class="p">,</span> <span class="n">n_probe</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">8</span><span class="p">,</span> <span class="n">n_train</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">50_000</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>加载数据集文本文件</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">43</span> <span class="n">dataset</span> <span class="o">=</span> <span class="n">TextFileDataset</span><span class="p">(</span>
|
||||
<span class="lineno">44</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">'tiny_shakespeare.txt'</span><span class="p">,</span>
|
||||
<span class="lineno">45</span> <span class="nb">list</span><span class="p">,</span>
|
||||
<span class="lineno">46</span> <span class="n">url</span><span class="o">=</span><span class="s1">'https://raw.githubusercontent.com/karpathy/char-rnn/master/data/tinyshakespeare/input.txt'</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>获取训练数据（字符串）</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">49</span> <span class="n">text</span> <span class="o">=</span> <span class="n">dataset</span><span class="o">.</span><span class="n">train</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>将文本分成大块<code class="highlight"><span></span><span class="n">chunk_length</span></code>
|
||||
</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">52</span> <span class="n">chunks</span> <span class="o">=</span> <span class="p">[</span><span class="n">text</span><span class="p">[</span><span class="n">i</span><span class="p">:</span><span class="n">i</span> <span class="o">+</span> <span class="n">chunk_len</span><span class="p">]</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="mi">0</span><span class="p">,</span> <span class="nb">len</span><span class="p">(</span><span class="n">text</span><span class="p">),</span> <span class="n">chunk_len</span><span class="p">)</span> <span class="k">if</span> <span class="n">i</span> <span class="o">+</span> <span class="n">chunk_len</span> <span class="o">*</span> <span class="mi">2</span> <span class="o"><</span> <span class="nb">len</span><span class="p">(</span><span class="n">text</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>获取每个区块的偏移量</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">54</span> <span class="n">chunk_offsets</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">i</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="mi">0</span><span class="p">,</span> <span class="nb">len</span><span class="p">(</span><span class="n">text</span><span class="p">),</span> <span class="n">chunk_len</span><span class="p">)</span> <span class="k">if</span> <span class="n">i</span> <span class="o">+</span> <span class="n">chunk_len</span> <span class="o">*</span> <span class="mi">2</span> <span class="o"><</span> <span class="nb">len</span><span class="p">(</span><span class="n">text</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>区块数</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">56</span> <span class="n">n_chunks</span> <span class="o">=</span> <span class="nb">len</span><span class="p">(</span><span class="n">chunks</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 BERT 以获取<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 eqa" style=""><span class="mord text" style=""><span class="mord" style="">B</span><span class="mord sizing reset-size6 size5" style=""><span class="mord" style="">ERT</span></span></span><span class="mopen" style="">(</span><span class="mord mathnormal" style="margin-right:0.10903em">N</span><span class="mclose" style="">)</span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">59</span> <span class="n">bert</span> <span class="o">=</span> <span class="n">BERTChunkEmbeddings</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">device</span><span class="p">(</span><span class="s1">'cuda:0'</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>通过处理每次迭代的块<code class="highlight"><span></span><span class="n">batch_size</span></code>
|
||||
数来获取区块嵌入</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">62</span> <span class="n">chunk_emb</span> <span class="o">=</span> <span class="p">[]</span>
|
||||
<span class="lineno">63</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">'Get embeddings'</span><span class="p">,</span> <span class="nb">range</span><span class="p">(</span><span class="mi">0</span><span class="p">,</span> <span class="n">n_chunks</span><span class="p">,</span> <span class="n">batch_size</span><span class="p">)):</span>
|
||||
<span class="lineno">64</span> <span class="n">chunk_emb</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">bert</span><span class="p">(</span><span class="n">chunks</span><span class="p">[</span><span class="n">i</span><span class="p">:</span> <span class="n">i</span> <span class="o">+</span> <span class="n">batch_size</span><span class="p">])</span><span class="o">.</span><span class="n">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>将它们合并成单个张量</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">66</span> <span class="n">chunk_emb</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">chunk_emb</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="o">.</span><span class="n">numpy</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>创建 <a href="https://faiss.ai/cpp_api/struct/structfaiss_1_1IndexIVFPQ.html">FAISS 指数</a></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">69</span> <span class="n">quantizer</span> <span class="o">=</span> <span class="n">faiss</span><span class="o">.</span><span class="n">IndexFlatL2</span><span class="p">(</span><span class="n">d_emb</span><span class="p">)</span>
|
||||
<span class="lineno">70</span> <span class="n">index</span> <span class="o">=</span> <span class="n">faiss</span><span class="o">.</span><span class="n">IndexIVFPQ</span><span class="p">(</span><span class="n">quantizer</span><span class="p">,</span> <span class="n">d_emb</span><span class="p">,</span> <span class="n">n_centeroids</span><span class="p">,</span> <span class="n">code_size</span><span class="p">,</span> <span class="mi">8</span><span class="p">)</span>
|
||||
<span class="lineno">71</span> <span class="n">index</span><span class="o">.</span><span class="n">nprobe</span> <span class="o">=</span> <span class="n">n_probe</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>获取区块索引的随机样本</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">74</span> <span class="n">random_sample</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">choice</span><span class="p">(</span><span class="n">np</span><span class="o">.</span><span class="n">arange</span><span class="p">(</span><span class="n">n_chunks</span><span class="p">),</span> <span class="n">size</span><span class="o">=</span><span class="p">[</span><span class="nb">min</span><span class="p">(</span><span class="n">n_train</span><span class="p">,</span> <span class="n">n_chunks</span><span class="p">)],</span> <span class="n">replace</span><span class="o">=</span><span class="kc">False</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>训练索引以存储密钥</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">77</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">'Train index'</span><span class="p">):</span>
|
||||
<span class="lineno">78</span> <span class="n">index</span><span class="o">.</span><span class="n">train</span><span class="p">(</span><span class="n">chunk_emb</span><span class="p">[</span><span class="n">random_sample</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>将块按大小批次添加到索引中<code class="highlight"><span></span><span class="mi">1024</span></code>
|
||||
</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">81</span> <span class="k">for</span> <span class="n">s</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">'Index'</span><span class="p">,</span> <span class="nb">range</span><span class="p">(</span><span class="mi">0</span><span class="p">,</span> <span class="n">n_chunks</span><span class="p">,</span> <span class="mi">1024</span><span class="p">)):</span>
|
||||
<span class="lineno">82</span> <span class="n">e</span> <span class="o">=</span> <span class="nb">min</span><span class="p">(</span><span class="n">s</span> <span class="o">+</span> <span class="mi">1024</span><span class="p">,</span> <span class="n">n_chunks</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>添加到索引</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">84</span> <span class="n">index</span><span class="o">.</span><span class="n">add_with_ids</span><span class="p">(</span><span class="n">chunk_emb</span><span class="p">[</span><span class="n">s</span><span class="p">:</span><span class="n">e</span><span class="p">],</span> <span class="n">chunk_offsets</span><span class="p">[</span><span class="n">s</span><span class="p">:</span> <span class="n">e</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>保存索引</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">87</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">'Save'</span><span class="p">):</span>
|
||||
<span class="lineno">88</span> <span class="n">faiss</span><span class="o">.</span><span class="n">write_index</span><span class="p">(</span><span class="n">index</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="o">/</span> <span class="s1">'retro.index'</span><span class="p">))</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>
|
||||
<h2>检索最近邻居的索引</h2>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">91</span><span class="k">class</span> <span class="nc">RetroIndex</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">chunk_len</span></code>
|
||||
是区块长度</li>
|
||||
<li><code class="highlight"><span></span><span class="n">n_probe</span></code>
|
||||
是要探测的列表的数量</li>
|
||||
<li><code class="highlight"><span></span><span class="n">n_neighbors</span></code>
|
||||
是要检索的邻居数量</li>
|
||||
<li><code class="highlight"><span></span><span class="n">n_extra</span></code>
|
||||
是要检索的额外邻居的数量，因为我们将移除与查询块重叠的邻居</li>
|
||||
<li><code class="highlight"><span></span><span class="n">exclude_neighbor_span</span></code>
|
||||
是检查重叠时要避免的额外文本长度</li></ul>
|
||||
|
||||
</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">chunk_len</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">16</span><span class="p">,</span> <span class="n">n_probe</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">8</span><span class="p">,</span>
|
||||
<span class="lineno">97</span> <span class="n">n_neighbors</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">2</span><span class="p">,</span> <span class="n">n_extra</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">2</span><span class="p">,</span>
|
||||
<span class="lineno">98</span> <span class="n">exclude_neighbor_span</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">8</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">108</span> <span class="bp">self</span><span class="o">.</span><span class="n">n_neighbors</span> <span class="o">=</span> <span class="n">n_neighbors</span>
|
||||
<span class="lineno">109</span> <span class="bp">self</span><span class="o">.</span><span class="n">chunk_len</span> <span class="o">=</span> <span class="n">chunk_len</span>
|
||||
<span class="lineno">110</span> <span class="bp">self</span><span class="o">.</span><span class="n">exclude_neighbor_span</span> <span class="o">=</span> <span class="n">exclude_neighbor_span</span>
|
||||
<span class="lineno">111</span> <span class="bp">self</span><span class="o">.</span><span class="n">n_extra</span> <span class="o">=</span> <span class="n">n_extra</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 BERT 以获取<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 eqa" style=""><span class="mord text" style=""><span class="mord" style="">B</span><span class="mord sizing reset-size6 size5" style=""><span class="mord" style="">ERT</span></span></span><span class="mopen" style="">(</span><span class="mord mathnormal" style="margin-right:0.10903em">N</span><span class="mclose" style="">)</span></span></span></span></span></span></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">114</span> <span class="bp">self</span><span class="o">.</span><span class="n">bert</span> <span class="o">=</span> <span class="n">BERTChunkEmbeddings</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">device</span><span class="p">(</span><span class="s1">'cuda:0'</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>装载数据库</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">116</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 index'</span><span class="p">):</span>
|
||||
<span class="lineno">117</span> <span class="bp">self</span><span class="o">.</span><span class="n">index</span> <span class="o">=</span> <span class="n">faiss</span><span class="o">.</span><span class="n">read_index</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="o">/</span> <span class="s1">'retro.index'</span><span class="p">))</span>
|
||||
<span class="lineno">118</span> <span class="bp">self</span><span class="o">.</span><span class="n">index</span><span class="o">.</span><span class="n">nprobe</span> <span class="o">=</span> <span class="n">n_probe</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>
|
||||
<h4>筛选与查询重叠的邻域</h4>
|
||||
<p>邻居的位置由给出<code class="highlight"><span></span><span class="n">neighbor_offsets</span></code>
|
||||
，查询块的位置为<code class="highlight"><span></span><span class="n">offset</span></code>
|
||||
。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">120</span> <span class="k">def</span> <span class="nf">filter_neighbors</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">offset</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">neighbor_offsets</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-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">127</span> <span class="k">return</span> <span class="p">[</span><span class="n">n</span> <span class="k">for</span> <span class="n">n</span> <span class="ow">in</span> <span class="n">neighbor_offsets</span>
|
||||
<span class="lineno">128</span> <span class="k">if</span> <span class="n">n</span> <span class="o"><</span> <span class="n">offset</span> <span class="o">-</span> <span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">chunk_len</span> <span class="o">+</span> <span class="bp">self</span><span class="o">.</span><span class="n">exclude_neighbor_span</span><span class="p">)</span>
|
||||
<span class="lineno">129</span> <span class="ow">or</span> <span class="n">n</span> <span class="o">></span> <span class="n">offset</span> <span class="o">+</span> <span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">chunk_len</span> <span class="o">+</span> <span class="bp">self</span><span class="o">.</span><span class="n">exclude_neighbor_span</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>检索最近的邻居</h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">131</span> <span class="k">def</span> <span class="fm">__call__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">query_chunks</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="n">offsets</span><span class="p">:</span> <span class="n">Optional</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-24'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-24'>#</a>
|
||||
</div>
|
||||
<p>获取<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 eqa" style=""><span class="mord text" style=""><span class="mord" style="">B</span><span class="mord sizing reset-size6 size5" style=""><span class="mord" style="">ERT</span></span></span><span class="mopen" style="">(</span><span class="mord mathnormal" style="margin-right:0.10903em">N</span><span class="mclose" style="">)</span></span></span></span></span></span>查询区块</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">137</span> <span class="n">emb</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">bert</span><span class="p">(</span><span class="n">query_chunks</span><span class="p">)</span><span class="o">.</span><span class="n">cpu</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>从数据库中获取<code class="highlight"><span></span><span class="n">n_neighbors</span> <span class="o">+</span> <span class="n">n_extra</span></code>
|
||||
最近的邻居</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">140</span> <span class="n">distance</span><span class="p">,</span> <span class="n">neighbor_offsets</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">index</span><span class="o">.</span><span class="n">search</span><span class="p">(</span><span class="n">emb</span><span class="o">.</span><span class="n">numpy</span><span class="p">(),</span> <span class="bp">self</span><span class="o">.</span><span class="n">n_neighbors</span> <span class="o">+</span> <span class="bp">self</span><span class="o">.</span><span class="n">n_extra</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>如果给出了查询区块偏移量，请过滤掉重叠的块</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">143</span> <span class="k">if</span> <span class="n">offsets</span> <span class="ow">is</span> <span class="ow">not</span> <span class="kc">None</span><span class="p">:</span>
|
||||
<span class="lineno">144</span> <span class="n">neighbor_offsets</span> <span class="o">=</span> <span class="p">[</span><span class="bp">self</span><span class="o">.</span><span class="n">filter_neighbors</span><span class="p">(</span><span class="n">off</span><span class="p">,</span> <span class="n">n_off</span><span class="p">)</span>
|
||||
<span class="lineno">145</span> <span class="k">for</span> <span class="n">off</span><span class="p">,</span> <span class="n">n_off</span> <span class="ow">in</span> <span class="nb">zip</span><span class="p">(</span><span class="n">offsets</span><span class="p">,</span> <span class="n">neighbor_offsets</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>筛选<code class="highlight"><span></span><span class="n">n_neighbors</span></code>
|
||||
后获取最接近的值</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">148</span> <span class="n">neighbor_offsets</span> <span class="o">=</span> <span class="p">[</span><span class="n">n_off</span><span class="p">[:</span><span class="bp">self</span><span class="o">.</span><span class="n">n_neighbors</span><span class="p">]</span> <span class="k">for</span> <span class="n">n_off</span> <span class="ow">in</span> <span class="n">neighbor_offsets</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></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">151</span> <span class="k">return</span> <span class="n">neighbor_offsets</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></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">155</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">156</span> <span class="n">build_database</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 +1,530 @@
|
||||
<!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="创建用于 RETRO 模型训练的数据集"/>
|
||||
|
||||
<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="RETRO 的训练数据集"/>
|
||||
<meta name="twitter:description" content="创建用于 RETRO 模型训练的数据集"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/transformers/retro/dataset.html"/>
|
||||
<meta property="og:title" content="RETRO 的训练数据集"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="RETRO 的训练数据集"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="RETRO 的训练数据集"/>
|
||||
<meta property="og:description" content="创建用于 RETRO 模型训练的数据集"/>
|
||||
|
||||
<title>RETRO 的训练数据集</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/transformers/retro/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">transformers</a>
|
||||
<a class="parent" href="index.html">retro</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/transformers/retro/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>
|
||||
<h1>RETRO 训练数据集</h1>
|
||||
<p>我们从<a href="database.html">键值数据库</a>中预先检索最近的邻域，并创建数据集来训练 <a href="index.html">RETRO</a> <a href="model.html">模型</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">json</span>
|
||||
<span class="lineno">16</span><span class="kn">from</span> <span class="nn">pathlib</span> <span class="kn">import</span> <span class="n">Path</span>
|
||||
<span class="lineno">17</span>
|
||||
<span class="lineno">18</span><span class="kn">import</span> <span class="nn">numpy</span> <span class="k">as</span> <span class="nn">np</span>
|
||||
<span class="lineno">19</span><span class="kn">import</span> <span class="nn">torch</span>
|
||||
<span class="lineno">20</span><span class="kn">from</span> <span class="nn">torch.utils.data</span> <span class="kn">import</span> <span class="n">Dataset</span> <span class="k">as</span> <span class="n">PyTorchDataset</span>
|
||||
<span class="lineno">21</span>
|
||||
<span class="lineno">22</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">monit</span>
|
||||
<span class="lineno">23</span><span class="kn">from</span> <span class="nn">labml_helpers.datasets.text</span> <span class="kn">import</span> <span class="n">TextFileDataset</span><span class="p">,</span> <span class="n">TextDataset</span>
|
||||
<span class="lineno">24</span><span class="kn">from</span> <span class="nn">labml_nn.transformers.retro.database</span> <span class="kn">import</span> <span class="n">RetroIndex</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>构建数据集</h2>
|
||||
<ul><li><code class="highlight"><span></span><span class="n">chunk_len</span></code>
|
||||
是区块长度</li>
|
||||
<li><code class="highlight"><span></span><span class="n">chunks_per_sample</span></code>
|
||||
是每个训练样本的块数</li>
|
||||
<li><code class="highlight"><span></span><span class="n">skip_range</span></code>
|
||||
是在两个样本之间跳过的最大字符数。我们在样本之间跳过几个字符，以确保样本与<a href="database.html">数据库</a>中的块不完全对齐</li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">27</span><span class="k">def</span> <span class="nf">build_dataset</span><span class="p">(</span><span class="n">chunk_len</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">16</span><span class="p">,</span> <span class="n">chunks_per_sample</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">32</span><span class="p">,</span> <span class="n">skip_range</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">8</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>加载文本文件</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">39</span> <span class="n">dataset</span> <span class="o">=</span> <span class="n">TextFileDataset</span><span class="p">(</span>
|
||||
<span class="lineno">40</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">'tiny_shakespeare.txt'</span><span class="p">,</span>
|
||||
<span class="lineno">41</span> <span class="nb">list</span><span class="p">,</span>
|
||||
<span class="lineno">42</span> <span class="n">url</span><span class="o">=</span><span class="s1">'https://raw.githubusercontent.com/karpathy/char-rnn/master/data/tinyshakespeare/input.txt'</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>其中的一部分训练</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">45</span> <span class="n">text</span> <span class="o">=</span> <span class="n">dataset</span><span class="o">.</span><span class="n">train</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>加载索引以检索邻居</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">48</span> <span class="n">index</span> <span class="o">=</span> <span class="n">RetroIndex</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>输入样本偏移</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">51</span> <span class="n">sample_offsets</span> <span class="o">=</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>光标指向文本</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">53</span> <span class="n">i</span> <span class="o">=</span> <span class="mi">0</span>
|
||||
<span class="lineno">54</span> <span class="k">while</span> <span class="n">i</span> <span class="o"><</span> <span class="nb">len</span><span class="p">(</span><span class="n">text</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>跳过几个角色以确保它不与邻居对齐</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">56</span> <span class="n">skip</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="n">skip_range</span><span class="p">)</span>
|
||||
<span class="lineno">57</span> <span class="n">i</span> <span class="o">+=</span> <span class="n">skip</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>如果我们到达了文字的末尾，就停下来</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">60</span> <span class="k">if</span> <span class="n">i</span> <span class="o">+</span> <span class="n">chunks_per_sample</span> <span class="o">*</span> <span class="n">chunk_len</span> <span class="o">></span> <span class="nb">len</span><span class="p">(</span><span class="n">text</span><span class="p">):</span>
|
||||
<span class="lineno">61</span> <span class="k">break</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>收集偏移量</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">64</span> <span class="n">sample_offsets</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">i</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>增加光标</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">67</span> <span class="n">i</span> <span class="o">+=</span> <span class="n">chunks_per_sample</span> <span class="o">*</span> <span class="n">chunk_len</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>对于样品</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">70</span> <span class="n">samples</span> <span class="o">=</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>遍历样本偏移量</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">72</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">'Gather Neighbors'</span><span class="p">,</span> <span class="n">sample_offsets</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>获取包含额外字符的样本（用于预测）</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">74</span> <span class="n">sample</span> <span class="o">=</span> <span class="n">text</span><span class="p">[</span><span class="n">i</span><span class="p">:</span> <span class="n">i</span> <span class="o">+</span> <span class="n">chunks_per_sample</span> <span class="o">*</span> <span class="n">chunk_len</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'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-14'>#</a>
|
||||
</div>
|
||||
<p>输入</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">76</span> <span class="n">src</span> <span class="o">=</span> <span class="n">sample</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>把它分成大块</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">78</span> <span class="n">chunks</span> <span class="o">=</span> <span class="p">[</span><span class="n">src</span><span class="p">[</span><span class="n">j</span><span class="p">:</span><span class="n">j</span> <span class="o">+</span> <span class="n">chunk_len</span><span class="p">]</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="mi">0</span><span class="p">,</span> <span class="nb">len</span><span class="p">(</span><span class="n">src</span><span class="p">),</span> <span class="n">chunk_len</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>区块偏移量</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">80</span> <span class="n">chunk_offsets</span> <span class="o">=</span> <span class="p">[</span><span class="n">j</span> <span class="o">+</span> <span class="n">i</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="mi">0</span><span class="p">,</span> <span class="nb">len</span><span class="p">(</span><span class="n">src</span><span class="p">),</span> <span class="n">chunk_len</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>检索最近的邻居</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">83</span> <span class="n">neighbor_offsets</span> <span class="o">=</span> <span class="n">index</span><span class="p">(</span><span class="n">chunks</span><span class="p">,</span> <span class="n">chunk_offsets</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>获取邻居短信。邻居长度是两倍<code class="highlight"><span></span><span class="n">chunk_len</span></code>
|
||||
</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">86</span> <span class="n">neighbors</span> <span class="o">=</span> <span class="p">[[</span><span class="n">text</span><span class="p">[</span><span class="n">j</span><span class="p">:</span> <span class="n">j</span> <span class="o">+</span> <span class="n">chunk_len</span> <span class="o">*</span> <span class="mi">2</span><span class="p">]</span> <span class="k">for</span> <span class="n">j</span> <span class="ow">in</span> <span class="n">n_off</span><span class="p">]</span> <span class="k">for</span> <span class="n">n_off</span> <span class="ow">in</span> <span class="n">neighbor_offsets</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>添加到样品清单</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">89</span> <span class="n">samples</span><span class="o">.</span><span class="n">append</span><span class="p">((</span><span class="n">sample</span><span class="p">[:</span><span class="o">-</span><span class="mi">1</span><span class="p">],</span> <span class="n">sample</span><span class="p">[</span><span class="mi">1</span><span class="p">:],</span> <span class="n">neighbors</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>以 JSON 格式保存示例。我们不需要使用复杂的数据集存储机制或预标记化，因为我们的数据集很小。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">94</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">lab</span><span class="o">.</span><span class="n">get_data_path</span><span class="p">()</span> <span class="o">/</span> <span class="s1">'retro_train_dataset.json'</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">95</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">samples</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>
|
||||
<h2>数据集</h2>
|
||||
<p>这是 PyTorch 数据集，用于加载由创建的数据集<code class="highlight"><span></span><span class="n">build_dataset</span></code>
|
||||
。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">98</span><span class="k">class</span> <span class="nc">Dataset</span><span class="p">(</span><span class="n">PyTorchDataset</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">file_path</span></code>
|
||||
是保存的 JSON 文件的路径</li>
|
||||
</ul><li><code class="highlight"><span></span><span class="n">tds</span></code>
|
||||
是<code class="highlight"><span></span><span class="n">TextDataset</span></code>
|
||||
</li>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">105</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">file_path</span><span class="p">:</span> <span class="n">Path</span><span class="p">,</span> <span class="n">tds</span><span class="p">:</span> <span class="n">TextDataset</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">111</span> <span class="bp">self</span><span class="o">.</span><span class="n">tds</span> <span class="o">=</span> <span class="n">tds</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>加载样品</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">113</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">file_path</span><span class="p">),</span> <span class="s1">'r'</span><span class="p">)</span> <span class="k">as</span> <span class="n">f</span><span class="p">:</span>
|
||||
<span class="lineno">114</span> <span class="bp">self</span><span class="o">.</span><span class="n">samples</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></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-25'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-25'>#</a>
|
||||
</div>
|
||||
<p>样本数量</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">116</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-26'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-26'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">120</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">samples</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>
|
||||
<p>获取样品</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">122</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-28'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-28'>#</a>
|
||||
</div>
|
||||
<p>获取样品</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">127</span> <span class="n">s</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">samples</span><span class="p">[</span><span class="n">idx</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>Tokenize</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">129</span> <span class="n">src</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">tds</span><span class="o">.</span><span class="n">text_to_i</span><span class="p">(</span><span class="n">s</span><span class="p">[</span><span class="mi">0</span><span class="p">])</span>
|
||||
<span class="lineno">130</span> <span class="n">tgt</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">tds</span><span class="o">.</span><span class="n">text_to_i</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="lineno">131</span> <span class="n">neighbors</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">stack</span><span class="p">([</span><span class="n">torch</span><span class="o">.</span><span class="n">stack</span><span class="p">([</span><span class="bp">self</span><span class="o">.</span><span class="n">tds</span><span class="o">.</span><span class="n">text_to_i</span><span class="p">(</span><span class="n">n</span><span class="p">)</span> <span class="k">for</span> <span class="n">n</span> <span class="ow">in</span> <span class="n">chunks</span><span class="p">])</span> <span class="k">for</span> <span class="n">chunks</span> <span class="ow">in</span> <span class="n">s</span><span class="p">[</span><span class="mi">2</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></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">133</span> <span class="k">return</span> <span class="n">src</span><span class="p">,</span> <span class="n">tgt</span><span class="p">,</span> <span class="n">neighbors</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></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">136</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">137</span> <span class="n">build_dataset</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 +1,136 @@
|
||||
<!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="这是论文《通过从数万亿个令牌中检索改进语言模型》的 PyTorch 实现/教程。它构建了一个包含文本块的键值数据库，并在进行预测时检索和使用它们。"/>
|
||||
|
||||
<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="检索增强型变压器（复古）"/>
|
||||
<meta name="twitter:description" content="这是论文《通过从数万亿个令牌中检索改进语言模型》的 PyTorch 实现/教程。它构建了一个包含文本块的键值数据库，并在进行预测时检索和使用它们。"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/transformers/retro/index.html"/>
|
||||
<meta property="og:title" content="检索增强型变压器（复古）"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="检索增强型变压器（复古）"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="检索增强型变压器（复古）"/>
|
||||
<meta property="og:description" content="这是论文《通过从数万亿个令牌中检索改进语言模型》的 PyTorch 实现/教程。它构建了一个包含文本块的键值数据库，并在进行预测时检索和使用它们。"/>
|
||||
|
||||
<title>检索增强型变压器（复古）</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/transformers/retro/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">transformers</a>
|
||||
<a class="parent" href="index.html">retro</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/transformers/retro/__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>检索增强型变压器（复古）</h1>
|
||||
<p>这是 <a href="https://pytorch.org">PyTorch</a> 对论文《<a href="https://arxiv.org/abs/2112.04426">通过从数万亿个代币中检索来改进语言模型》的</a>实现。</p>
|
||||
<p>它建立了一个包含大量文本的数据库。它是一个键值数据库，其中的密钥由区块的 BERT 嵌入索引。他们使用冻结的预训练的 BERT 模型来计算这些嵌入。这些值是相应的区块和该区块的等长度文本。</p>
|
||||
<p>然后，模型从该数据库检索与模型输入相似（最近邻域）的文本。这些检索到的文本用于预测输出。</p>
|
||||
<p>由于我们使用冻结的 BERT 模型进行检索，因此我们可以预先计算训练数据集的所有最近邻域。这加快了训练过程。</p>
|
||||
<p>组件：</p>
|
||||
<ul><li><a href="bert_embeddings.html">BERT 嵌入</a>：用于获取大块文本的 BERT 嵌入的代码。</li>
|
||||
<li><a href="database.html">键值数据库</a>：生成和检索区块</li>
|
||||
<li><a href="model.html">模型</a></li>
|
||||
<li><a href="dataset.html">数据集</a>：预先计算最近的邻居</li>
|
||||
<li><a href="train.html">训练代码</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 +1,855 @@
|
||||
<!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="使用小莎士比亚数据集训练 RETRO 模型"/>
|
||||
|
||||
<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="复古训练"/>
|
||||
<meta name="twitter:description" content="使用小莎士比亚数据集训练 RETRO 模型"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/transformers/retro/train.html"/>
|
||||
<meta property="og:title" content="复古训练"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="复古训练"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="复古训练"/>
|
||||
<meta property="og:description" content="使用小莎士比亚数据集训练 RETRO 模型"/>
|
||||
|
||||
<title>复古训练</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/transformers/retro/train.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">transformers</a>
|
||||
<a class="parent" href="index.html">retro</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/transformers/retro/train.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>复古训练</h1>
|
||||
<p>这是 RETRO 的训练<a href="index.html">代码</a>。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">14</span><span></span><span class="kn">import</span> <span class="nn">torch</span>
|
||||
<span class="lineno">15</span><span class="kn">from</span> <span class="nn">torch</span> <span class="kn">import</span> <span class="n">nn</span>
|
||||
<span class="lineno">16</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">RandomSampler</span>
|
||||
<span class="lineno">17</span>
|
||||
<span class="lineno">18</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">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">logger</span>
|
||||
<span class="lineno">19</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">20</span><span class="kn">from</span> <span class="nn">labml_helpers.datasets.text</span> <span class="kn">import</span> <span class="n">TextFileDataset</span>
|
||||
<span class="lineno">21</span><span class="kn">from</span> <span class="nn">labml_nn.optimizers.noam</span> <span class="kn">import</span> <span class="n">Noam</span>
|
||||
<span class="lineno">22</span><span class="kn">from</span> <span class="nn">labml_nn.transformers.retro</span> <span class="kn">import</span> <span class="n">model</span> <span class="k">as</span> <span class="n">retro</span>
|
||||
<span class="lineno">23</span><span class="kn">from</span> <span class="nn">labml_nn.transformers.retro.dataset</span> <span class="kn">import</span> <span class="n">Dataset</span><span class="p">,</span> <span class="n">RetroIndex</span>
|
||||
<span class="lineno">24</span><span class="kn">from</span> <span class="nn">labml_nn.transformers.retro.model</span> <span class="kn">import</span> <span class="n">RetroModel</span><span class="p">,</span> <span class="n">NearestNeighborEncoder</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>采样器</h2>
|
||||
<p>这个类贪婪地从模型中抽样。</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">27</span><span class="k">class</span> <span class="nc">Sampler</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">device</span></code>
|
||||
是该型号的设备</li>
|
||||
<li><code class="highlight"><span></span><span class="n">model</span></code>
|
||||
是复<a href="retro.html">古模式</a></li>
|
||||
<li><code class="highlight"><span></span><span class="n">tds</span></code>
|
||||
是文本数据集（用于获取相邻数据块）</li>
|
||||
<li><code class="highlight"><span></span><span class="n">chunk_len</span></code>
|
||||
是区块的长度</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">device</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">device</span><span class="p">,</span> <span class="n">model</span><span class="p">:</span> <span class="n">retro</span><span class="o">.</span><span class="n">RetroModel</span><span class="p">,</span> <span class="n">tds</span><span class="p">:</span> <span class="n">TextFileDataset</span><span class="p">,</span> <span class="n">chunk_len</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">chunk_len</span> <span class="o">=</span> <span class="n">chunk_len</span>
|
||||
<span class="lineno">42</span> <span class="bp">self</span><span class="o">.</span><span class="n">tds</span> <span class="o">=</span> <span class="n">tds</span>
|
||||
<span class="lineno">43</span> <span class="bp">self</span><span class="o">.</span><span class="n">model</span> <span class="o">=</span> <span class="n">model</span>
|
||||
<span class="lineno">44</span> <span class="bp">self</span><span class="o">.</span><span class="n">device</span> <span class="o">=</span> <span class="n">device</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="database.html">复古指数</a></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">index</span> <span class="o">=</span> <span class="n">RetroIndex</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>检索给定区块的最近邻域</h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">49</span> <span class="k">def</span> <span class="nf">retrieve_nearest_neighbours</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">chunk</span><span class="p">:</span> <span class="nb">str</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>检索最近邻点的偏移量</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">55</span> <span class="n">neighbor_offsets</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">index</span><span class="p">([</span><span class="n">chunk</span><span class="p">],</span> <span class="kc">None</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>获取邻居（邻居长度等于<code class="highlight"><span></span><span class="n">chunk_len</span> <span class="o">*</span> <span class="mi">2</span></code>
|
||||
）</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">58</span> <span class="n">text</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">tds</span><span class="o">.</span><span class="n">train</span>
|
||||
<span class="lineno">59</span> <span class="n">neighbors</span> <span class="o">=</span> <span class="p">[</span><span class="n">text</span><span class="p">[</span><span class="n">j</span><span class="p">:</span> <span class="n">j</span> <span class="o">+</span> <span class="bp">self</span><span class="o">.</span><span class="n">chunk_len</span> <span class="o">*</span> <span class="mi">2</span><span class="p">]</span> <span class="k">for</span> <span class="n">j</span> <span class="ow">in</span> <span class="n">neighbor_offsets</span><span class="p">[</span><span class="mi">0</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></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">62</span> <span class="k">return</span> <span class="n">neighbors</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>来自给定提示的示例文本</h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">64</span> <span class="k">def</span> <span class="nf">sample</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">prompt</span><span class="p">:</span> <span class="nb">str</span><span class="p">,</span> <span class="n">sample_len</span><span class="p">:</span> <span class="nb">int</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>将最近邻存储为字符串</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">70</span> <span class="n">neighbors_str</span> <span class="o">=</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>样本文本</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">73</span> <span class="n">sampled</span> <span class="o">=</span> <span class="s1">''</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><code class="highlight"><span></span><span class="n">sample_len</span></code>
|
||||
代币样本</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">76</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">sample_len</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>如果采样块比我们已经检索到的要多，我们就需要检索邻居</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">79</span> <span class="k">while</span> <span class="nb">len</span><span class="p">(</span><span class="n">neighbors_str</span><span class="p">)</span> <span class="o"><</span> <span class="nb">len</span><span class="p">(</span><span class="n">prompt</span><span class="p">)</span> <span class="o">//</span> <span class="bp">self</span><span class="o">.</span><span class="n">chunk_len</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>获取我们尚未检索到邻居的最后一个区块</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">81</span> <span class="n">off</span> <span class="o">=</span> <span class="nb">len</span><span class="p">(</span><span class="n">neighbors_str</span><span class="p">)</span> <span class="o">*</span> <span class="bp">self</span><span class="o">.</span><span class="n">chunk_len</span>
|
||||
<span class="lineno">82</span> <span class="n">chunk</span> <span class="o">=</span> <span class="n">prompt</span><span class="p">[</span><span class="n">off</span><span class="p">:</span> <span class="n">off</span> <span class="o">+</span> <span class="bp">self</span><span class="o">.</span><span class="n">chunk_len</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>检索最近的邻居</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">84</span> <span class="n">neighbors_str</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">retrieve_nearest_neighbours</span><span class="p">(</span><span class="n">chunk</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>对输入进行标记化</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">87</span> <span class="n">src</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">tds</span><span class="o">.</span><span class="n">text_to_i</span><span class="p">(</span><span class="n">prompt</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>标记检索到的邻居</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">89</span> <span class="n">neighbors</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">stack</span><span class="p">([</span><span class="n">torch</span><span class="o">.</span><span class="n">stack</span><span class="p">([</span><span class="bp">self</span><span class="o">.</span><span class="n">tds</span><span class="o">.</span><span class="n">text_to_i</span><span class="p">(</span><span class="n">n</span><span class="p">)</span> <span class="k">for</span> <span class="n">n</span> <span class="ow">in</span> <span class="n">chunk</span><span class="p">])</span> <span class="k">for</span> <span class="n">chunk</span> <span class="ow">in</span> <span class="n">neighbors_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>
|
||||
<p>将它们移到与模型相同的设备上</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">92</span> <span class="n">src</span> <span class="o">=</span> <span class="n">src</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="lineno">93</span> <span class="n">neighbors</span> <span class="o">=</span> <span class="n">neighbors</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-19'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-19'>#</a>
|
||||
</div>
|
||||
<p>获取模型输出</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">96</span> <span class="n">res</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">src</span><span class="p">[</span><span class="kc">None</span><span class="p">,</span> <span class="p">:],</span> <span class="n">neighbors</span><span class="p">[</span><span class="kc">None</span><span class="p">,</span> <span class="p">:,</span> <span class="p">:,</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>贪婪地抽取最后一个代币</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">99</span> <span class="n">token</span> <span class="o">=</span> <span class="n">res</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="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-21'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-21'>#</a>
|
||||
</div>
|
||||
<p>将采样标记文本添加到提示和示例文本中</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">102</span> <span class="n">prompt</span> <span class="o">+=</span> <span class="bp">self</span><span class="o">.</span><span class="n">tds</span><span class="o">.</span><span class="n">itos</span><span class="p">[</span><span class="n">token</span><span class="o">.</span><span class="n">item</span><span class="p">()]</span>
|
||||
<span class="lineno">103</span> <span class="n">sampled</span> <span class="o">+=</span> <span class="bp">self</span><span class="o">.</span><span class="n">tds</span><span class="o">.</span><span class="n">itos</span><span class="p">[</span><span class="n">token</span><span class="o">.</span><span class="n">item</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></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">106</span> <span class="k">return</span> <span class="n">sampled</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>
|
||||
<h2>复古教练</h2>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">109</span><span class="k">class</span> <span class="nc">Trainer</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>
|
||||
<ul><li><code class="highlight"><span></span><span class="n">device</span></code>
|
||||
是该型号的设备</li>
|
||||
<li><code class="highlight"><span></span><span class="n">model</span></code>
|
||||
是复<a href="retro.html">古模式</a></li>
|
||||
<li><code class="highlight"><span></span><span class="n">dataloader</span></code>
|
||||
是<a href="dataset.html">具有预检索邻域的数据集的数据</a>加载器</li>
|
||||
<li><code class="highlight"><span></span><span class="n">optimizer</span></code>
|
||||
是优化器</li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">114</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">device</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">device</span><span class="p">,</span> <span class="n">model</span><span class="p">:</span> <span class="n">retro</span><span class="o">.</span><span class="n">RetroModel</span><span class="p">,</span>
|
||||
<span class="lineno">115</span> <span class="n">dataloader</span><span class="p">:</span> <span class="n">DataLoader</span><span class="p">,</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">Optimizer</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">122</span> <span class="bp">self</span><span class="o">.</span><span class="n">optimizer</span> <span class="o">=</span> <span class="n">optimizer</span>
|
||||
<span class="lineno">123</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">124</span> <span class="bp">self</span><span class="o">.</span><span class="n">dataloader</span> <span class="o">=</span> <span class="n">dataloader</span>
|
||||
<span class="lineno">125</span> <span class="bp">self</span><span class="o">.</span><span class="n">model</span> <span class="o">=</span> <span class="n">model</span>
|
||||
<span class="lineno">126</span> <span class="bp">self</span><span class="o">.</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-26'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-26'>#</a>
|
||||
</div>
|
||||
<h3>为一个时代训练模型</h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">128</span> <span class="k">def</span> <span class="fm">__call__</span><span class="p">(</span><span class="bp">self</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>遍历训练数据</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">134</span> <span class="k">for</span> <span class="n">i</span><span class="p">,</span> <span class="p">(</span><span class="n">src</span><span class="p">,</span> <span class="n">tgt</span><span class="p">,</span> <span class="n">neighbors</span><span class="p">)</span> <span class="ow">in</span> <span class="n">monit</span><span class="o">.</span><span class="n">enum</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">dataloader</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>将数据移动到设备</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">136</span> <span class="n">src</span><span class="p">,</span> <span class="n">tgt</span><span class="p">,</span> <span class="n">neighbors</span> <span class="o">=</span> <span class="n">src</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">tgt</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">neighbors</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-29'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-29'>#</a>
|
||||
</div>
|
||||
<p>向前传球</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">139</span> <span class="n">res</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">src</span><span class="p">,</span> <span class="n">neighbors</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>计算损失</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">141</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">res</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="n">res</span><span class="o">.</span><span class="n">shape</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">]),</span> <span class="n">tgt</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></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-31'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-31'>#</a>
|
||||
</div>
|
||||
<p>清除渐变</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">144</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-32'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-32'>#</a>
|
||||
</div>
|
||||
<p>向后传球</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">146</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-33'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-33'>#</a>
|
||||
</div>
|
||||
<p>优化模型</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">148</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-34'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-34'>#</a>
|
||||
</div>
|
||||
<p>保存训练统计数据并增加全局步数计数器</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">151</span> <span class="n">tracker</span><span class="o">.</span><span class="n">save</span><span class="p">({</span><span class="s1">'loss.train'</span><span class="p">:</span> <span class="n">loss</span><span class="p">})</span>
|
||||
<span class="lineno">152</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">src</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>
|
||||
<h2>创建和训练一个小模型</h2>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">155</span><span class="k">def</span> <span class="nf">train</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>创建实验</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">161</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">'retro_small'</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>GPU 设备</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">164</span> <span class="n">device</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">device</span><span class="p">(</span><span class="s1">'cuda: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>加载小莎士比亚数据集</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">167</span> <span class="n">tds</span> <span class="o">=</span> <span class="n">TextFileDataset</span><span class="p">(</span>
|
||||
<span class="lineno">168</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">'tiny_shakespeare.txt'</span><span class="p">,</span>
|
||||
<span class="lineno">169</span> <span class="nb">list</span><span class="p">,</span>
|
||||
<span class="lineno">170</span> <span class="n">url</span><span class="o">=</span><span class="s1">'https://raw.githubusercontent.com/karpathy/char-rnn/master/data/tinyshakespeare/input.txt'</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>加载<a href="dataset.html">复古数据集</a></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">173</span> <span class="n">train_dataset</span> <span class="o">=</span> <span class="n">Dataset</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="o">/</span> <span class="s1">'retro_train_dataset.json'</span><span class="p">,</span> <span class="n">tds</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>创建数据加载器</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">176</span> <span class="n">train_dl</span> <span class="o">=</span> <span class="n">DataLoader</span><span class="p">(</span><span class="n">train_dataset</span><span class="p">,</span>
|
||||
<span class="lineno">177</span> <span class="n">batch_size</span><span class="o">=</span><span class="mi">4</span><span class="p">,</span>
|
||||
<span class="lineno">178</span> <span class="n">sampler</span><span class="o">=</span><span class="n">RandomSampler</span><span class="p">(</span><span class="n">train_dataset</span><span class="p">,</span> <span class="n">replacement</span><span class="o">=</span><span class="kc">True</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>超参数</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">181</span> <span class="n">chunk_len</span> <span class="o">=</span> <span class="mi">16</span>
|
||||
<span class="lineno">182</span> <span class="n">d_model</span> <span class="o">=</span> <span class="mi">128</span>
|
||||
<span class="lineno">183</span> <span class="n">d_ff</span> <span class="o">=</span> <span class="mi">512</span>
|
||||
<span class="lineno">184</span> <span class="n">n_heads</span> <span class="o">=</span> <span class="mi">16</span>
|
||||
<span class="lineno">185</span> <span class="n">d_k</span> <span class="o">=</span> <span class="mi">16</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>创建最近邻编码器</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">188</span> <span class="n">nearest_neighbor_encoder</span> <span class="o">=</span> <span class="n">NearestNeighborEncoder</span><span class="p">(</span><span class="n">chunk_len</span><span class="p">,</span> <span class="mi">6</span><span class="p">,</span> <span class="p">{</span><span class="mi">3</span><span class="p">},</span> <span class="n">d_model</span><span class="p">,</span> <span class="n">n_heads</span><span class="p">,</span> <span class="n">d_k</span><span class="p">,</span> <span class="n">d_ff</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>创建模型</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">190</span> <span class="n">model</span> <span class="o">=</span> <span class="n">RetroModel</span><span class="p">(</span><span class="n">tds</span><span class="o">.</span><span class="n">n_tokens</span><span class="p">,</span> <span class="n">d_model</span><span class="p">,</span> <span class="mi">6</span><span class="p">,</span>
|
||||
<span class="lineno">191</span> <span class="p">{</span><span class="mi">3</span><span class="p">,</span> <span class="mi">5</span><span class="p">},</span>
|
||||
<span class="lineno">192</span> <span class="n">chunk_len</span><span class="p">,</span> <span class="n">n_heads</span><span class="p">,</span> <span class="n">d_k</span><span class="p">,</span> <span class="n">d_ff</span><span class="p">,</span>
|
||||
<span class="lineno">193</span> <span class="n">encoder</span><span class="o">=</span><span class="n">nearest_neighbor_encoder</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>将模型移到设备上</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">195</span> <span class="n">model</span> <span class="o">=</span> <span class="n">model</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">device</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>创建优化器</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">197</span> <span class="n">optimizer</span> <span class="o">=</span> <span class="n">Noam</span><span class="p">(</span><span class="n">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="mf">1.</span><span class="p">,</span> <span class="n">d_model</span><span class="o">=</span><span class="n">d_model</span><span class="p">,</span> <span class="n">warmup</span><span class="o">=</span><span class="mi">2_000</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<code class="highlight"><span></span><span class="n">Trainer</span></code>
|
||||
</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">199</span> <span class="n">trainer</span> <span class="o">=</span> <span class="n">Trainer</span><span class="p">(</span><span class="n">device</span><span class="p">,</span> <span class="n">model</span><span class="p">,</span> <span class="n">train_dl</span><span class="p">,</span> <span class="n">optimizer</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>创建<code class="highlight"><span></span><span class="n">Sampler</span></code>
|
||||
</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">201</span> <span class="n">sampler</span> <span class="o">=</span> <span class="n">Sampler</span><span class="p">(</span><span class="n">device</span><span class="p">,</span> <span class="n">model</span><span class="p">,</span> <span class="n">tds</span><span class="p">,</span> <span class="n">chunk_len</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></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">203</span> <span class="n">prompt</span> <span class="o">=</span> <span class="s1">'''Second Citizen:</span><span class="se">\n</span><span class="s1">One word, good citizens.</span><span class="se">\n\n</span><span class="s1">First Citizen:'''</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>设置用于保存和加载的模型</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">206</span> <span class="n">experiment</span><span class="o">.</span><span class="n">add_pytorch_models</span><span class="p">(</span><span class="n">model</span><span class="o">=</span><span class="n">model</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>开始实验</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">209</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-51'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-51'>#</a>
|
||||
</div>
|
||||
<p>为<code class="highlight"><span></span><span class="mi">32</span></code>
|
||||
时代而训练</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">211</span> <span class="k">for</span> <span class="n">epoch</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="mi">32</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>火车</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">213</span> <span class="n">trainer</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>打印一条新行</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">215</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-54'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-54'>#</a>
|
||||
</div>
|
||||
<p>样本来自<code class="highlight"><span></span><span class="n">prompt</span></code>
|
||||
</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">217</span> <span class="n">logger</span><span class="o">.</span><span class="n">log</span><span class="p">([(</span><span class="n">prompt</span><span class="o">.</span><span class="n">replace</span><span class="p">(</span><span class="s1">'</span><span class="se">\n</span><span class="s1">'</span><span class="p">,</span> <span class="s1">'</span><span class="se">\\</span><span class="s1">n</span><span class="se">\n</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="lineno">218</span> <span class="p">(</span><span class="n">sampler</span><span class="o">.</span><span class="n">sample</span><span class="p">(</span><span class="n">prompt</span><span class="p">,</span> <span class="mi">128</span><span class="p">)</span><span class="o">.</span><span class="n">replace</span><span class="p">(</span><span class="s1">'</span><span class="se">\n</span><span class="s1">'</span><span class="p">,</span> <span class="s1">'</span><span class="se">\\</span><span class="s1">n</span><span class="se">\n</span><span class="s1">'</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-55'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-55'>#</a>
|
||||
</div>
|
||||
<p>保存模型</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">220</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-56'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-56'>#</a>
|
||||
</div>
|
||||
<p></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">224</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">225</span> <span class="n">train</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>
|
||||
Reference in New Issue
Block a user