chore: import upstream snapshot with attribution
This commit is contained in:
wehub-resource-sync
@@ -0,0 +1,522 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="ja">
|
||||
<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="フリーズした BERT モデルを使用してチャンクの BERT 埋め込みを生成"/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="BERT テキストの塊の埋め込み"/>
|
||||
<meta name="twitter:description" content="フリーズした BERT モデルを使用してチャンクの BERT 埋め込みを生成"/>
|
||||
<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 テキストの塊の埋め込み"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="BERT テキストの塊の埋め込み"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="BERT テキストの塊の埋め込み"/>
|
||||
<meta property="og:description" content="フリーズした BERT モデルを使用してチャンクの BERT 埋め込みを生成"/>
|
||||
|
||||
<title>BERT テキストの塊の埋め込み</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 テキストの塊の埋め込み</h1>
|
||||
<p><a href="index.html">これは、RETROモデル用のチャンクのBERT埋め込みを取得するためのコードです。</a></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">13</span><span></span><span class="kn">from</span> <span class="nn">typing</span> <span class="kn">import</span> <span class="n">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 エンベディング</h2>
|
||||
<p><span ><span class="katex"><span aria-hidden="true" class="katex-html"><span class="base"><span class="strut" style="height:0.68333em;vertical-align:0em;"></span><span class="mord coloredeq eqb" style=""><span class="mord mathnormal" style="margin-right:0.10903em">N</span></span></span></span></span></span>このクラスは、特定のテキストチャンクに対して BERT <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> 埋め込みを生成します。<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>は、内のすべてのトークンのBERT埋め込みの平均です。<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><a href="https://huggingface.co/bert-base-uncased">HuggingFace から BERT トークナイザーをロードします</a></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><a href="https://huggingface.co/bert-base-uncased">ハギングフェイスから</a> BERT モデルをロード</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>モデルを次の場所に移動 <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>この実装では、固定数のトークンでチャンクを作成しません。理由の1つは、この実装では文字レベルのトークンを使用し、BERTはサブワードトークナイザーを使用していることです</p>。
|
||||
<p>そのため、このメソッドはテキストを切り捨てて、部分的なトークンがないことを確認します。</p>
|
||||
<p>たとえば、チャンクは、末尾に単語の一部（部分的なサブワードトークン）<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>
|
||||
があるようなものかもしれません。BERT 埋め込みの精度を高めるため、これらを削除しました。先に述べたように、トークン化後にチャンクを分割した場合は必要ありません</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>空白を削除</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>ブレーク・ワード</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>最初と最後のピースを削除する</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>空白を削除</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>空の場合、元の文字列を返す</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>それ以外の場合は、取り除いた文字列を返す</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>チャンクのリストを入手してください。</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>勾配を計算する必要はありません</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>チャンクをトリミング</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>BERT トークナイザーでチャンクをトークン化する</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>トークン ID、アテンションマスク、トークンタイプをデバイスに移動</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>モデルの評価</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>トークンの埋め込みを入手</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>トークン埋め込みの平均回数を計算します。注意マスクは、<code class="highlight"><span></span><span class="mi">0</span></code>
|
||||
トークンが空でパディングされている場合であることに注意してください。チャンクの長さが異なるため、空のトークンが返されます</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>BERT 埋め込みをテストするコード</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>[初期化]</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>[サンプル]</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>BERT トークナイザーをチェック</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>BERT モデルの出力をチェック</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>トークン ID からテキストを再作成することを確認してください</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>チャンク埋め込みを取得</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://papers.labml.ai">Trending Research Papers</a>
|
||||
<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 +1,536 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="ja">
|
||||
<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="最近傍検索とデータベースの作成"/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="最近傍検索用データベース"/>
|
||||
<meta name="twitter:description" content="最近傍検索とデータベースの作成"/>
|
||||
<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="最近傍検索用データベース"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="最近傍検索用データベース"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="最近傍検索用データベース"/>
|
||||
<meta property="og:description" content="最近傍検索とデータベースの作成"/>
|
||||
|
||||
<title>最近傍検索用データベース</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>最近傍検索用データベース</h1>
|
||||
<p>これは、<a href="index.html">RETROモデルのデータベースを構築し</a>、最も近い近傍を検索するものです。</p>
|
||||
</a><p>論文ではScanNライブラリを使用していましたが、<a href="https://faiss.ai/">データベースにはFAISSライブラリを使用しています。</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>データベースを構築</h2>
|
||||
<ul><li><code class="highlight"><span></span><span class="n">chunk_len</span></code>
|
||||
チャンクの長さ (文字数)</li>
|
||||
<li><code class="highlight"><span></span><span class="n">batch_size</span></code>
|
||||
計算時に使用するバッチサイズです <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>
|
||||
FAISS <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> <a href="https://faiss.ai/cpp_api/struct/structfaiss_1_1IndexIVFPQ.html">インデックスで選択する埋め込みリスト内のフィーチャの数です</a></li>
|
||||
<li><code class="highlight"><span></span><span class="n">n_centeroids</span></code>
|
||||
インデックス内のリストの数です</li>
|
||||
<li><code class="highlight"><span></span><span class="n">code_size</span></code>
|
||||
インデックス内のエンコードされたベクトルサイズ</li>
|
||||
<li><code class="highlight"><span></span><span class="n">n_probe</span></code>
|
||||
は調べるリストの数です</li>
|
||||
<li>`n_train' はインデックスをトレーニングするキーの数です</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>データセットのテキストファイルを読み込む</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>トレーニングデータ (文字列) を取得</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>テキストを次のチャンクに分割します <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>各チャンクのオフセットを取得</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>チャンクの数</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>BERT を初期化して取得 <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><code class="highlight"><span></span><span class="n">batch_size</span></code>
|
||||
各反復処理でチャンクの数を処理してチャンクの埋め込みを行う</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>それらを単一のテンソルにマージします</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><a href="https://faiss.ai/cpp_api/struct/structfaiss_1_1IndexIVFPQ.html">FAISS</a> インデックスを作成する</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>チャンクインデックスのランダムサンプルを取得</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>キーを保存するようにインデックスをトレーニングする</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>チャンクをサイズごとにインデックスに追加します。<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>索引に追加</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>インデックスを保存する</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>最近傍を検索するためのインデックス</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>
|
||||
はチャンクの長さです</li>
|
||||
<li><code class="highlight"><span></span><span class="n">n_probe</span></code>
|
||||
は調べるリストの数です</li>
|
||||
<li><code class="highlight"><span></span><span class="n">n_neighbors</span></code>
|
||||
取得する近傍の数です</li>
|
||||
<li><code class="highlight"><span></span><span class="n">n_extra</span></code>
|
||||
クエリチャンクと重複している近傍を削除することになるため、取得する余分な近傍の数です</li>
|
||||
<li><code class="highlight"><span></span><span class="n">exclude_neighbor_span</span></code>
|
||||
オーバーラップをチェックするときに避けるべき余分なテキスト長です</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>BERT を初期化して取得 <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>データベースを読み込む</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>クエリと重複する近傍をフィルタリングする</h4>
|
||||
<p><code class="highlight"><span></span><span class="n">neighbor_offsets</span></code>
|
||||
近傍の位置はで指定され、クエリチャンクの位置はです。<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>最も近い近傍を検索する</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><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>クエリチャンクの取得</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><code class="highlight"><span></span><span class="n">n_neighbors</span> <span class="o">+</span> <span class="n">n_extra</span></code>
|
||||
データベースから最も近い近傍を取得</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>クエリのチャンクオフセットが指定されている場合は、重複するチャンクを除外します。</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><code class="highlight"><span></span><span class="n">n_neighbors</span></code>
|
||||
フィルタリング後に最も近いものを取得</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://papers.labml.ai">Trending Research Papers</a>
|
||||
<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 +1,531 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="ja">
|
||||
<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="RETRO モデルトレーニング用のデータセットの作成"/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="RETRO 用トレーニングデータセット"/>
|
||||
<meta name="twitter:description" content="RETRO モデルトレーニング用のデータセットの作成"/>
|
||||
<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 用トレーニングデータセット"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="RETRO 用トレーニングデータセット"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="RETRO 用トレーニングデータセット"/>
|
||||
<meta property="og:description" content="RETRO モデルトレーニング用のデータセットの作成"/>
|
||||
|
||||
<title>RETRO 用トレーニングデータセット</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 トレーニングデータセット</h1>
|
||||
<p><a href="database.html">キーバリューデータベースから最も近い近傍を事前に取得し</a><a href="index.html"><a href="model.html">、RETRO モデルをトレーニングするためのデータセットを作成します。</a></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>データセットの構築</h2>
|
||||
<ul><li><code class="highlight"><span></span><span class="n">chunk_len</span></code>
|
||||
はチャンクの長さです</li>
|
||||
<li><code class="highlight"><span></span><span class="n">chunks_per_sample</span></code>
|
||||
はトレーニングサンプルあたりのチャンク数です</li>
|
||||
<li><code class="highlight"><span></span><span class="n">skip_range</span></code>
|
||||
2 つのサンプル間でスキップする最大文字数です。</li></ul><a href="database.html">サンプルがデータベース内のチャンクと完全に一致していないことを確認するために、サンプル間を数文字スキップしています。</a>
|
||||
|
||||
</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>テキストファイルを読み込む</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>そのトレーニング部分</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>近傍検索用のインデックスを読み込む</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>入力サンプルオフセット</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>テキスト用のカーソル</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>数文字飛ばして、隣の文字と揃わないようにしてください</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>テキストの終わりに達したら止めてください</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>オフセットの収集</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>カーソルをインクリメントしてください</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>サンプル用</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>サンプルオフセットを反復処理</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>追加文字を含むサンプルを取得 (予測用)</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>インプット</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>それをチャンクに分割してください</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>チャンクオフセット</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>最も近い近傍を検索する</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>近所の人のテキストを取得.近傍の長さは 2 倍です</p>。<code class="highlight"><span></span><span class="n">chunk_len</span></code>
|
||||
|
||||
|
||||
</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>サンプルリストに追加</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>サンプルを JSON で保存します。データセットは小さいため、複雑なデータセットの保存メカニズムを使用したり、事前にトークン化したりする必要はありません</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>データセット</h2>
|
||||
<p>これは、によって作成されたデータセットをロードする PyTorch データセットです。<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>
|
||||
保存された JSON ファイルのパスです</li>
|
||||
</ul><li><code class="highlight"><span></span><span class="n">tds</span></code>
|
||||
は <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>サンプルを読み込む</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>サンプル数</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>サンプルを入手</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>サンプルを入手</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>トークン化</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://papers.labml.ai">Trending Research Papers</a>
|
||||
<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 +1,137 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="ja">
|
||||
<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="これは、数兆のトークンから取得することによる言語モデルの改善という論文のPyTorch実装/チュートリアルです。テキストのチャンクから成るキーバリューデータベースを構築し、それらを取得して予測に使用します。"/>
|
||||
|
||||
<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="検索機能付きトランスフォーマー (レトロ)"/>
|
||||
<meta name="twitter:description" content="これは、数兆のトークンから取得することによる言語モデルの改善という論文のPyTorch実装/チュートリアルです。テキストのチャンクから成るキーバリューデータベースを構築し、それらを取得して予測に使用します。"/>
|
||||
<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="検索機能付きトランスフォーマー (レトロ)"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="検索機能付きトランスフォーマー (レトロ)"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="検索機能付きトランスフォーマー (レトロ)"/>
|
||||
<meta property="og:description" content="これは、数兆のトークンから取得することによる言語モデルの改善という論文のPyTorch実装/チュートリアルです。テキストのチャンクから成るキーバリューデータベースを構築し、それらを取得して予測に使用します。"/>
|
||||
|
||||
<title>検索機能付きトランスフォーマー (レトロ)</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>検索機能付きトランスフォーマー (レトロ)</h1>
|
||||
<p>これは、<a href="https://pytorch.org"><a href="https://papers.labml.ai/paper/2112.04426">数兆のトークンから取得することによる言語モデルの改善という論文をPyTorchで実装したものです</a></a>。</p>
|
||||
<p>テキストの塊のデータベースを構築します。これは、チャンクのBERT埋め込みによってキーのインデックスが付けられるキーバリューデータベースです。これらの埋め込みを計算するには、事前にトレーニング済みのフリーズした BERT モデルを使用します。値は、対応するチャンクと、そのチャンクの前に続く同じ長さのテキストです。</p>
|
||||
<p>次に、モデルはこのデータベースからモデルへの入力に類似した (最も近い近傍の) テキストを取得します。取得したこれらのテキストは、出力の予測に使用されます</p>。
|
||||
<p>検索にはフリーズした BERT モデルを使用するため、トレーニングデータセットのすべての最近傍を事前に計算できます。これにより、トレーニングプロセスがスピードアップします。</p>
|
||||
<p>コンポーネント:</p>
|
||||
<ul><li><a href="bert_embeddings.html">BERT 埋め込み:テキストのチャンクを</a> BERT に埋め込むためのコード。</li>
|
||||
<li><a href="database.html">キーバリューデータベース:チャンクの構築と取得</a></li>
|
||||
<li><a href="model.html">モデル</a></li>
|
||||
<li><a href="dataset.html">データセット</a>:最も近い近傍を事前計算</li>
|
||||
<li><a href="train.html">トレーニングコード</a></li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://papers.labml.ai">Trending Research Papers</a>
|
||||
<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 +1,856 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="ja">
|
||||
<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="タイニーシェイクスピアデータセットによる RETRO モデルのトレーニング"/>
|
||||
|
||||
<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="レトロトレーニング"/>
|
||||
<meta name="twitter:description" content="タイニーシェイクスピアデータセットによる RETRO モデルのトレーニング"/>
|
||||
<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="レトロトレーニング"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="レトロトレーニング"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="レトロトレーニング"/>
|
||||
<meta property="og:description" content="タイニーシェイクスピアデータセットによる RETRO モデルのトレーニング"/>
|
||||
|
||||
<title>レトロトレーニング</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>レトロトレーニング</h1>
|
||||
<p><a href="index.html">これはRETROのトレーニングコードです</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>サンプラー</h2>
|
||||
<p>このクラスはモデルから貪欲にサンプリングします。</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>
|
||||
モデルのデバイスです</li>
|
||||
<li><code class="highlight"><span></span><span class="n">model</span></code>
|
||||
<a href="retro.html">レトロモードです</a></li>
|
||||
<li><code class="highlight"><span></span><span class="n">tds</span></code>
|
||||
テキストデータセット (隣接チャンクの取得に使用)</li>
|
||||
<li><code class="highlight"><span></span><span class="n">chunk_len</span></code>
|
||||
チャンクの長さです</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">レトロインデックス</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>与えられたチャンクの最も近い近傍のデータを取得</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>最も近い近傍のオフセットを取得</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>近傍を取得 (近傍の長さがと等しい) <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>与えられたプロンプトのサンプルテキスト</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>最も近い近傍を文字列として保存するには</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>サンプルテキスト</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>
|
||||
サンプルトークン</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>すでに取得したチャンクの数よりもサンプリングされたチャンクの数が多い場合は、隣接データを取得する必要があります</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>隣接データを取得していない最後のチャンクを取得</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>最も近い近傍を検索する</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>入力をトークン化</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>取得したネイバーをトークン化する</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>モデルと同じデバイスに移動します</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>モデル出力を取得</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>最後のトークンを欲張ってサンプリングする</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>サンプリングしたトークンのテキストをプロンプトとサンプルテキストに追加します</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>レトロトレーナー</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>
|
||||
モデルのデバイスです</li>
|
||||
<li><code class="highlight"><span></span><span class="n">model</span></code>
|
||||
<a href="retro.html">レトロモードです</a></li>
|
||||
<li><code class="highlight"><span></span><span class="n">dataloader</span></code>
|
||||
は、<a href="dataset.html">事前に近傍が取得されたデータセットのデータローダー</a></li>
|
||||
<li><code class="highlight"><span></span><span class="n">optimizer</span></code>
|
||||
オプティマイザーです</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>一時代を先取りしたモデルのトレーニング</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>トレーニングデータを繰り返し処理</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>データをデバイスに移動</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>フォワードパス</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>損失の計算</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>グラデーションをクリア</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>バックワードパス</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>モデルを最適化</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>トレーニング統計を保存してグローバルステップカウンタを増やす</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>小型モデルの作成とトレーニング</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>テストを作成</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 デバイス</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>タイニーシェイクスピアデータセットの読み込み</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><a href="dataset.html">レトロデータセットの読み込み</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>データローダーの作成</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>ハイパーパラメータ</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>最近傍エンコーダーの作成</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>モデルを作成</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>モデルをデバイスに移動</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>オプティマイザーの作成</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>作成 <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>作成 <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>保存および読み込み用のモデルを設定する</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>実験を始める</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><code class="highlight"><span></span><span class="mi">32</span></code>
|
||||
時代に合わせた列車</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>列車</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>新しい行を印刷</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>からのサンプル <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>モデルを保存する</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://papers.labml.ai">Trending Research Papers</a>
|
||||
<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