chore: import upstream snapshot with attribution
This commit is contained in:
wehub-resource-sync
@@ -0,0 +1,443 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="en">
|
||||
<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="This is training code with notes for a feedback transformer."/>
|
||||
|
||||
<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="Train Feedback Transformer"/>
|
||||
<meta name="twitter:description" content="This is training code with notes for a feedback transformer."/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/transformers/feedback/experiment.html"/>
|
||||
<meta property="og:title" content="Train Feedback Transformer"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="Train Feedback Transformer"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="Train Feedback Transformer"/>
|
||||
<meta property="og:description" content="This is training code with notes for a feedback transformer."/>
|
||||
|
||||
<title>Train Feedback Transformer</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/feedback/experiment.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="../index.html">transformers</a>
|
||||
<a class="parent" href="index.html">feedback</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/feedback/experiment.py" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1>Train Feedback Transformer</h1>
|
||||
<p>This trains a <a href="index.html">feedback transformer</a> model for auto-regression. You can pick the original feedback transformer or the new version where the keys and values are precalculated.</p>
|
||||
<p>Here's a Colab notebook for training a feedback transformer on Tiny Shakespeare dataset.</p>
|
||||
<p><a href="https://colab.research.google.com/github/labmlai/annotated_deep_learning_paper_implementations/blob/master/labml_nn/transformers/feedback/experiment.ipynb"><img alt="Open In Colab" src="https://colab.research.google.com/assets/colab-badge.svg"></a></p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">18</span><span></span><span class="kn">import</span> <span class="nn">torch</span>
|
||||
<span class="lineno">19</span><span class="kn">from</span> <span class="nn">labml</span> <span class="kn">import</span> <span class="n">experiment</span>
|
||||
<span class="lineno">20</span><span class="kn">from</span> <span class="nn">labml.configs</span> <span class="kn">import</span> <span class="n">option</span>
|
||||
<span class="lineno">21</span><span class="kn">from</span> <span class="nn">labml.utils.pytorch</span> <span class="kn">import</span> <span class="n">get_modules</span>
|
||||
<span class="lineno">22</span><span class="kn">from</span> <span class="nn">labml_nn.experiments.nlp_autoregression</span> <span class="kn">import</span> <span class="n">NLPAutoRegressionConfigs</span>
|
||||
<span class="lineno">23</span><span class="kn">from</span> <span class="nn">torch</span> <span class="kn">import</span> <span class="n">nn</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-1'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-1'>#</a>
|
||||
</div>
|
||||
<h2>Auto regressive model</h2>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">26</span><span class="k">class</span> <span class="nc">AutoregressiveModel</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-2'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-2'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">31</span> <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">n_vocab</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">d_model</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">transformer</span><span class="p">:</span> <span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span>
|
||||
<span class="lineno">32</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-3'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-3'>#</a>
|
||||
</div>
|
||||
<p>Token embedding module </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">34</span> <span class="bp">self</span><span class="o">.</span><span class="n">src_embed</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Embedding</span><span class="p">(</span><span class="n">n_vocab</span><span class="p">,</span> <span class="n">d_model</span><span class="p">)</span>
|
||||
<span class="lineno">35</span> <span class="bp">self</span><span class="o">.</span><span class="n">transformer</span> <span class="o">=</span> <span class="n">transformer</span>
|
||||
<span class="lineno">36</span> <span class="bp">self</span><span class="o">.</span><span class="n">generator</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Linear</span><span class="p">(</span><span class="n">d_model</span><span class="p">,</span> <span class="n">n_vocab</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-4'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-4'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">38</span> <span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-5'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-5'>#</a>
|
||||
</div>
|
||||
<p>Embed the tokens </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">40</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">src_embed</span><span class="p">(</span><span class="n">x</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-6'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-6'>#</a>
|
||||
</div>
|
||||
<p>Run it through the the transformer </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">42</span> <span class="n">res</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">transformer</span><span class="p">(</span><span class="n">x</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>Generate logits of the next token </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">44</span> <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">generator</span><span class="p">(</span><span class="n">res</span><span class="p">),</span> <span class="kc">None</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-8'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-8'>#</a>
|
||||
</div>
|
||||
<h2>Configurations</h2>
|
||||
<p>The default configs can and will be over-ridden when we start the experiment</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">47</span><span class="k">class</span> <span class="nc">Configs</span><span class="p">(</span><span class="n">NLPAutoRegressionConfigs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-9'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-9'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">54</span> <span class="n">model</span><span class="p">:</span> <span class="n">AutoregressiveModel</span>
|
||||
<span class="lineno">55</span>
|
||||
<span class="lineno">56</span> <span class="n">d_model</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">512</span>
|
||||
<span class="lineno">57</span> <span class="n">heads</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">8</span>
|
||||
<span class="lineno">58</span> <span class="n">dropout</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">0.0</span>
|
||||
<span class="lineno">59</span> <span class="n">d_ff</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">2048</span>
|
||||
<span class="lineno">60</span> <span class="n">n_layers</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">6</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-10'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-10'>#</a>
|
||||
</div>
|
||||
<p> Create <a href="index.html">original feedback transformer</a>.</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">63</span><span class="nd">@option</span><span class="p">(</span><span class="n">Configs</span><span class="o">.</span><span class="n">model</span><span class="p">)</span>
|
||||
<span class="lineno">64</span><span class="k">def</span> <span class="nf">feedback_transformer</span><span class="p">(</span><span class="n">c</span><span class="p">:</span> <span class="n">Configs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-11'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-11'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">68</span> <span class="kn">from</span> <span class="nn">labml_nn.transformers.feedback</span> <span class="kn">import</span> <span class="n">FeedbackTransformer</span><span class="p">,</span> <span class="n">FeedbackTransformerLayer</span><span class="p">,</span> \
|
||||
<span class="lineno">69</span> <span class="n">FeedbackAttention</span><span class="p">,</span> <span class="n">FeedForward</span>
|
||||
<span class="lineno">70</span>
|
||||
<span class="lineno">71</span> <span class="k">return</span> <span class="n">AutoregressiveModel</span><span class="p">(</span>
|
||||
<span class="lineno">72</span> <span class="n">c</span><span class="o">.</span><span class="n">n_tokens</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">d_model</span><span class="p">,</span>
|
||||
<span class="lineno">73</span> <span class="n">FeedbackTransformer</span><span class="p">(</span>
|
||||
<span class="lineno">74</span> <span class="n">FeedbackTransformerLayer</span><span class="p">(</span><span class="n">d_model</span><span class="o">=</span><span class="n">c</span><span class="o">.</span><span class="n">d_model</span><span class="p">,</span>
|
||||
<span class="lineno">75</span> <span class="n">attn</span><span class="o">=</span><span class="n">FeedbackAttention</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">heads</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">d_model</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">dropout</span><span class="p">),</span>
|
||||
<span class="lineno">76</span> <span class="n">feed_forward</span><span class="o">=</span><span class="n">FeedForward</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">d_model</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">d_ff</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">dropout</span><span class="p">),</span>
|
||||
<span class="lineno">77</span> <span class="n">dropout_prob</span><span class="o">=</span><span class="n">c</span><span class="o">.</span><span class="n">dropout</span><span class="p">),</span>
|
||||
<span class="lineno">78</span> <span class="n">c</span><span class="o">.</span><span class="n">n_layers</span><span class="p">))</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">device</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-12'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-12'>#</a>
|
||||
</div>
|
||||
<p> Create <a href="index.html#kv_shared">updated feedback transformer</a>, with precalculated keys and values.</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">81</span><span class="nd">@option</span><span class="p">(</span><span class="n">Configs</span><span class="o">.</span><span class="n">model</span><span class="p">)</span>
|
||||
<span class="lineno">82</span><span class="k">def</span> <span class="nf">feedback_transformer_kv</span><span class="p">(</span><span class="n">c</span><span class="p">:</span> <span class="n">Configs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-13'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-13'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">86</span> <span class="kn">from</span> <span class="nn">labml_nn.transformers.feedback</span> <span class="kn">import</span> <span class="n">FeedbackTransformerKV</span><span class="p">,</span> <span class="n">FeedbackTransformerLayer</span><span class="p">,</span> \
|
||||
<span class="lineno">87</span> <span class="n">FeedbackAttention</span><span class="p">,</span> <span class="n">FeedForward</span>
|
||||
<span class="lineno">88</span>
|
||||
<span class="lineno">89</span> <span class="k">return</span> <span class="n">AutoregressiveModel</span><span class="p">(</span>
|
||||
<span class="lineno">90</span> <span class="n">c</span><span class="o">.</span><span class="n">n_tokens</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">d_model</span><span class="p">,</span>
|
||||
<span class="lineno">91</span> <span class="n">FeedbackTransformerKV</span><span class="p">(</span>
|
||||
<span class="lineno">92</span> <span class="n">FeedbackTransformerLayer</span><span class="p">(</span><span class="n">d_model</span><span class="o">=</span><span class="n">c</span><span class="o">.</span><span class="n">d_model</span><span class="p">,</span>
|
||||
<span class="lineno">93</span> <span class="n">attn</span><span class="o">=</span><span class="n">FeedbackAttention</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">heads</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">d_model</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">dropout</span><span class="p">,</span>
|
||||
<span class="lineno">94</span> <span class="n">is_kv_precomputed</span><span class="o">=</span><span class="kc">True</span><span class="p">),</span>
|
||||
<span class="lineno">95</span> <span class="n">feed_forward</span><span class="o">=</span><span class="n">FeedForward</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">d_model</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">d_ff</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">dropout</span><span class="p">),</span>
|
||||
<span class="lineno">96</span> <span class="n">dropout_prob</span><span class="o">=</span><span class="n">c</span><span class="o">.</span><span class="n">dropout</span><span class="p">),</span>
|
||||
<span class="lineno">97</span> <span class="n">c</span><span class="o">.</span><span class="n">n_layers</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">d_model</span><span class="p">,</span> <span class="n">c</span><span class="o">.</span><span class="n">heads</span><span class="p">))</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">device</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-14'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-14'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">100</span><span class="k">def</span> <span class="nf">main</span><span class="p">():</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-15'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-15'>#</a>
|
||||
</div>
|
||||
<p>Create experiment </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">102</span> <span class="n">experiment</span><span class="o">.</span><span class="n">create</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s2">"feedback_transformer"</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>Create configs </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">104</span> <span class="n">conf</span> <span class="o">=</span> <span class="n">Configs</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-17'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-17'>#</a>
|
||||
</div>
|
||||
<p>Load configurations </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">106</span> <span class="n">experiment</span><span class="o">.</span><span class="n">configs</span><span class="p">(</span><span class="n">conf</span><span class="p">,</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-18'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-18'>#</a>
|
||||
</div>
|
||||
<p>A dictionary of configurations to override </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">108</span> <span class="p">{</span><span class="s1">'tokenizer'</span><span class="p">:</span> <span class="s1">'character'</span><span class="p">,</span>
|
||||
<span class="lineno">109</span> <span class="s1">'text'</span><span class="p">:</span> <span class="s1">'tiny_shakespeare'</span><span class="p">,</span>
|
||||
<span class="lineno">110</span> <span class="s1">'optimizer.learning_rate'</span><span class="p">:</span> <span class="mf">1.0</span><span class="p">,</span>
|
||||
<span class="lineno">111</span> <span class="s1">'optimizer.optimizer'</span><span class="p">:</span> <span class="s1">'Noam'</span><span class="p">,</span>
|
||||
<span class="lineno">112</span> <span class="s1">'prompt'</span><span class="p">:</span> <span class="s1">'It is'</span><span class="p">,</span>
|
||||
<span class="lineno">113</span> <span class="s1">'prompt_separator'</span><span class="p">:</span> <span class="s1">''</span><span class="p">,</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-19'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-19'>#</a>
|
||||
</div>
|
||||
<p>Use <code class="highlight"><span></span><span class="n">feedback_transformer</span></code>
|
||||
for original feedback transformer </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">116</span> <span class="s1">'model'</span><span class="p">:</span> <span class="s1">'feedback_transformer_kv'</span><span class="p">,</span>
|
||||
<span class="lineno">117</span>
|
||||
<span class="lineno">118</span> <span class="s1">'train_loader'</span><span class="p">:</span> <span class="s1">'shuffled_train_loader'</span><span class="p">,</span>
|
||||
<span class="lineno">119</span> <span class="s1">'valid_loader'</span><span class="p">:</span> <span class="s1">'shuffled_valid_loader'</span><span class="p">,</span>
|
||||
<span class="lineno">120</span>
|
||||
<span class="lineno">121</span> <span class="s1">'seq_len'</span><span class="p">:</span> <span class="mi">128</span><span class="p">,</span>
|
||||
<span class="lineno">122</span> <span class="s1">'epochs'</span><span class="p">:</span> <span class="mi">128</span><span class="p">,</span>
|
||||
<span class="lineno">123</span> <span class="s1">'batch_size'</span><span class="p">:</span> <span class="mi">64</span><span class="p">,</span>
|
||||
<span class="lineno">124</span> <span class="s1">'inner_iterations'</span><span class="p">:</span> <span class="mi">25</span><span class="p">})</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-20'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-20'>#</a>
|
||||
</div>
|
||||
<p>Set models for saving and loading </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">127</span> <span class="n">experiment</span><span class="o">.</span><span class="n">add_pytorch_models</span><span class="p">(</span><span class="n">get_modules</span><span class="p">(</span><span class="n">conf</span><span class="p">))</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-21'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-21'>#</a>
|
||||
</div>
|
||||
<p>Start the experiment </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">130</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-22'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-22'>#</a>
|
||||
</div>
|
||||
<p>Run the training loop </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">132</span> <span class="n">conf</span><span class="o">.</span><span class="n">run</span><span class="p">()</span>
|
||||
<span class="lineno">133</span>
|
||||
<span class="lineno">134</span>
|
||||
<span class="lineno">135</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">136</span> <span class="n">main</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
File diff suppressed because one or more lines are too long
@@ -0,0 +1,134 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="en">
|
||||
<head>
|
||||
<meta http-equiv="content-type" content="text/html;charset=utf-8"/>
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0"/>
|
||||
<meta name="description" content=""/>
|
||||
|
||||
<meta name="twitter:card" content="summary"/>
|
||||
<meta name="twitter:image:src" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta name="twitter:title" content="Feedback Transformer"/>
|
||||
<meta name="twitter:description" content=""/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/transformers/feedback/readme.html"/>
|
||||
<meta property="og:title" content="Feedback Transformer"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="Feedback Transformer"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="Feedback Transformer"/>
|
||||
<meta property="og:description" content=""/>
|
||||
|
||||
<title>Feedback Transformer</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/feedback/readme.html"/>
|
||||
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.18/dist/katex.min.css" integrity="sha384-zTROYFVGOfTw7JV7KUu8udsvW2fx4lWOsCEDqhBreBwlHI4ioVRtmIvEThzJHGET" crossorigin="anonymous">
|
||||
|
||||
<!-- Global site tag (gtag.js) - Google Analytics -->
|
||||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-4V3HC8HBLH"></script>
|
||||
<script>
|
||||
window.dataLayer = window.dataLayer || [];
|
||||
|
||||
function gtag() {
|
||||
dataLayer.push(arguments);
|
||||
}
|
||||
|
||||
gtag('js', new Date());
|
||||
|
||||
gtag('config', 'G-4V3HC8HBLH');
|
||||
</script>
|
||||
</head>
|
||||
<body>
|
||||
<div id='container'>
|
||||
<div id="background"></div>
|
||||
<div class='section'>
|
||||
<div class='docs'>
|
||||
<p>
|
||||
<a class="parent" href="/">home</a>
|
||||
<a class="parent" href="../index.html">transformers</a>
|
||||
<a class="parent" href="index.html">feedback</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/feedback/readme.md" target="_blank">
|
||||
View code on Github</a>
|
||||
</p>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1><a href="https://nn.labml.ai/transformers/feedback/index.html">Feedback Transformer</a></h1>
|
||||
<p>This is a <a href="https://pytorch.org">PyTorch</a> implementation of the paper <a href="https://arxiv.org/abs/2002.09402">Accessing Higher-level Representations in Sequential Transformers with Feedback Memory</a>.</p>
|
||||
<p>Normal transformers process tokens in parallel. Each transformer layer pays attention to the outputs of the previous layer. Feedback transformer pays attention to the output of all layers in previous steps. So this adds recurrence, and we need to process token-by-token. This slows down the training significantly (about 5X - 10X depending on the sequence length). However, when predicting Feedback Transformer is faster because you can predict the next token if you cache the memory vectors.</p>
|
||||
<p>In order to speed up the training the paper discusses starting with a short sequence length and gradually increasing it. They also discuss using a pretrained parallel transformer as the starting point.</p>
|
||||
<p>The original feedback transformer doesn't keep the outputs of all layers. Instead it keeps weighted sum of the output of all layers. This reduces the memory used for caching during prediction. The first half of this file implements this.</p>
|
||||
<p>The updated feedback transformer shares weights used to calculate keys and values among the layers. We then calculate the keys and values for each step only once and keep them cached. The <a href="#shared_kv">second half</a> of this file implements this. We implemented a custom PyTorch function to improve performance.</p>
|
||||
<p>Here's <a href="experiment.html">the training code</a> and a notebook for training a feedback transformer on Tiny Shakespeare dataset.</p>
|
||||
<p><a href="https://colab.research.google.com/github/labmlai/annotated_deep_learning_paper_implementations/blob/master/labml_nn/transformers/feedback/experiment.ipynb">Colab Notebook</a></p>
|
||||
<p><a href="https://colab.research.google.com/github/labmlai/annotated_deep_learning_paper_implementations/blob/master/labml_nn/transformers/feedback/experiment.ipynb"><img alt="Open In Colab" src="https://colab.research.google.com/assets/colab-badge.svg"></a> </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
<a href="https://labml.ai">labml.ai</a>
|
||||
</div>
|
||||
</div>
|
||||
<script src=../../interactive.js?v=1"></script>
|
||||
<script>
|
||||
function handleImages() {
|
||||
var images = document.querySelectorAll('p>img')
|
||||
|
||||
for (var i = 0; i < images.length; ++i) {
|
||||
handleImage(images[i])
|
||||
}
|
||||
}
|
||||
|
||||
function handleImage(img) {
|
||||
img.parentElement.style.textAlign = 'center'
|
||||
|
||||
var modal = document.createElement('div')
|
||||
modal.id = 'modal'
|
||||
|
||||
var modalContent = document.createElement('div')
|
||||
modal.appendChild(modalContent)
|
||||
|
||||
var modalImage = document.createElement('img')
|
||||
modalContent.appendChild(modalImage)
|
||||
|
||||
var span = document.createElement('span')
|
||||
span.classList.add('close')
|
||||
span.textContent = 'x'
|
||||
modal.appendChild(span)
|
||||
|
||||
img.onclick = function () {
|
||||
console.log('clicked')
|
||||
document.body.appendChild(modal)
|
||||
modalImage.src = img.src
|
||||
}
|
||||
|
||||
span.onclick = function () {
|
||||
document.body.removeChild(modal)
|
||||
}
|
||||
}
|
||||
|
||||
handleImages()
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
Reference in New Issue
Block a user