mirror of
https://github.com/labmlai/annotated_deep_learning_paper_implementations.git
synced 2025-10-29 01:26:44 +08:00
LoRA Chinese docs
This commit is contained in:
Varuna Jayasiri
@ -101,6 +101,7 @@
|
||||
<li><a href="transformers/vit/index.html">视觉 Transformer (ViT)</a></li>
|
||||
<li><a href="transformers/primer_ez/index.html">Primer</a></li>
|
||||
<li><a href="transformers/hour_glass/index.html">沙漏网络</a></li></ul>
|
||||
<h4>✨ <a href="lora/index.html">Low-Rank Adaptation (LoRA)</a></h4>
|
||||
<h4>✨ <a href="neox/index.html">Eleuther GPT-neox</a></h4>
|
||||
<ul><li><a href="neox/samples/generate.html">在一块 48GB GPU 上生成</a></li>
|
||||
<li><a href="neox/samples/finetune.html">在两块 48GB GPU 上微调</a></li>
|
||||
|
||||
@ -70,21 +70,22 @@
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
<h1>Finetune GPT-2 with <a href="index.html">LoRA</a></h1>
|
||||
<h1>Finetune <a href="gpt2.html">GPT-2</a> with <a href="index.html">LoRA</a></h1>
|
||||
<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/lora/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">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">labml</span> <span class="kn">import</span> <span class="n">lab</span><span class="p">,</span> <span class="n">monit</span><span class="p">,</span> <span class="n">tracker</span>
|
||||
<span class="lineno">16</span><span class="kn">from</span> <span class="nn">labml.configs</span> <span class="kn">import</span> <span class="n">BaseConfigs</span><span class="p">,</span> <span class="n">option</span>
|
||||
<span class="lineno">17</span><span class="kn">from</span> <span class="nn">labml.utils.download</span> <span class="kn">import</span> <span class="n">download_file</span>
|
||||
<span class="lineno">18</span><span class="kn">from</span> <span class="nn">labml_helpers.device</span> <span class="kn">import</span> <span class="n">DeviceConfigs</span>
|
||||
<span class="lineno">19</span><span class="kn">from</span> <span class="nn">torch.optim</span> <span class="kn">import</span> <span class="n">Adam</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">DataLoader</span><span class="p">,</span> <span class="n">TensorDataset</span>
|
||||
<span class="lineno">21</span><span class="kn">from</span> <span class="nn">transformers</span> <span class="kn">import</span> <span class="n">AutoTokenizer</span><span class="p">,</span> <span class="n">AutoModelForCausalLM</span>
|
||||
<span class="lineno">22</span><span class="kn">from</span> <span class="nn">labml_nn.lora.gpt2</span> <span class="kn">import</span> <span class="n">GPTModel</span></pre></div>
|
||||
<span class="lineno">15</span><span class="kn">from</span> <span class="nn">torch.optim</span> <span class="kn">import</span> <span class="n">Adam</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">TensorDataset</span>
|
||||
<span class="lineno">17</span><span class="kn">from</span> <span class="nn">transformers</span> <span class="kn">import</span> <span class="n">AutoTokenizer</span><span class="p">,</span> <span class="n">AutoModelForCausalLM</span>
|
||||
<span class="lineno">18</span>
|
||||
<span class="lineno">19</span><span class="kn">from</span> <span class="nn">labml</span> <span class="kn">import</span> <span class="n">lab</span><span class="p">,</span> <span class="n">monit</span><span class="p">,</span> <span class="n">tracker</span>
|
||||
<span class="lineno">20</span><span class="kn">from</span> <span class="nn">labml.configs</span> <span class="kn">import</span> <span class="n">BaseConfigs</span><span class="p">,</span> <span class="n">option</span>
|
||||
<span class="lineno">21</span><span class="kn">from</span> <span class="nn">labml.utils.download</span> <span class="kn">import</span> <span class="n">download_file</span>
|
||||
<span class="lineno">22</span><span class="kn">from</span> <span class="nn">labml_helpers.device</span> <span class="kn">import</span> <span class="n">DeviceConfigs</span>
|
||||
<span class="lineno">23</span><span class="kn">from</span> <span class="nn">labml_nn.lora.gpt2</span> <span class="kn">import</span> <span class="n">GPTModel</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-1'>
|
||||
@ -97,7 +98,7 @@
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">25</span><span class="k">class</span> <span class="nc">Trainer</span><span class="p">(</span><span class="n">BaseConfigs</span><span class="p">):</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">26</span><span class="k">class</span> <span class="nc">Trainer</span><span class="p">(</span><span class="n">BaseConfigs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-2'>
|
||||
@ -108,7 +109,7 @@
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">31</span> <span class="n">device</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">device</span> <span class="o">=</span> <span class="n">DeviceConfigs</span><span class="p">()</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">32</span> <span class="n">device</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">device</span> <span class="o">=</span> <span class="n">DeviceConfigs</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-3'>
|
||||
@ -120,11 +121,12 @@
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">34</span> <span class="n">layer_norm_epsilon</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">1e-05</span>
|
||||
<span class="lineno">35</span> <span class="n">n_embed</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">768</span>
|
||||
<span class="lineno">36</span> <span class="n">n_layer</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">12</span>
|
||||
<span class="lineno">37</span> <span class="n">n_positions</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">1024</span>
|
||||
<span class="lineno">38</span> <span class="n">vocab_size</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">50257</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">35</span> <span class="n">layer_norm_epsilon</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">1e-05</span>
|
||||
<span class="lineno">36</span> <span class="n">d_model</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">768</span>
|
||||
<span class="lineno">37</span> <span class="n">n_layers</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">12</span>
|
||||
<span class="lineno">38</span> <span class="n">n_heads</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">12</span>
|
||||
<span class="lineno">39</span> <span class="n">n_positions</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">1024</span>
|
||||
<span class="lineno">40</span> <span class="n">vocab_size</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">50257</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-4'>
|
||||
@ -136,10 +138,10 @@
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">41</span> <span class="n">epochs</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">10</span>
|
||||
<span class="lineno">42</span> <span class="n">batch_size</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">32</span>
|
||||
<span class="lineno">43</span> <span class="n">learning_rate</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">1e-4</span>
|
||||
<span class="lineno">44</span> <span class="n">context_len</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">512</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">43</span> <span class="n">epochs</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">10</span>
|
||||
<span class="lineno">44</span> <span class="n">batch_size</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">32</span>
|
||||
<span class="lineno">45</span> <span class="n">learning_rate</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">1e-4</span>
|
||||
<span class="lineno">46</span> <span class="n">context_len</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">512</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-5'>
|
||||
@ -151,7 +153,7 @@
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">47</span> <span class="n">lora_r</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">32</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">49</span> <span class="n">lora_r</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">32</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-6'>
|
||||
@ -163,24 +165,19 @@
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">50</span> <span class="n">text</span><span class="p">:</span> <span class="n">TensorDataset</span> <span class="o">=</span> <span class="s2">"tiny_shakespeare"</span>
|
||||
<span class="lineno">51</span> <span class="n">tokenizer</span> <span class="o">=</span> <span class="n">AutoTokenizer</span><span class="o">.</span><span class="n">from_pretrained</span><span class="p">(</span><span class="s2">"gpt2"</span><span class="p">)</span>
|
||||
<span class="lineno">52</span> <span class="n">model</span><span class="p">:</span> <span class="n">GPTModel</span>
|
||||
<span class="lineno">53</span> <span class="n">optimizer</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">optim</span><span class="o">.</span><span class="n">Adam</span>
|
||||
<span class="lineno">54</span> <span class="n">criterion</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">nn</span><span class="o">.</span><span class="n">CrossEntropyLoss</span><span class="p">()</span>
|
||||
<span class="lineno">55</span> <span class="n">data_loader</span><span class="p">:</span> <span class="n">DataLoader</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">52</span> <span class="n">text</span><span class="p">:</span> <span class="n">TensorDataset</span> <span class="o">=</span> <span class="s2">"tiny_shakespeare"</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-7'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-7'>#</a>
|
||||
</div>
|
||||
<h3>Load pre-trained <a href="https://huggingface.co/openai-community/gpt2">GPT-2 from huggingface</a></h3>
|
||||
<p>Huggingface tokenizer </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">57</span> <span class="k">def</span> <span class="nf">_load_pretrained_weights</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">54</span> <span class="n">tokenizer</span> <span class="o">=</span> <span class="n">AutoTokenizer</span><span class="o">.</span><span class="n">from_pretrained</span><span class="p">(</span><span class="s2">"gpt2"</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-8'>
|
||||
@ -188,12 +185,11 @@
|
||||
<div class='section-link'>
|
||||
<a href='#section-8'>#</a>
|
||||
</div>
|
||||
<p>Load the huggingface model and get the parameters </p>
|
||||
<p><a href="gpt2.html">GPT2 model</a> </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">63</span> <span class="n">hf_model</span> <span class="o">=</span> <span class="n">AutoModelForCausalLM</span><span class="o">.</span><span class="n">from_pretrained</span><span class="p">(</span><span class="s2">"gpt2"</span><span class="p">)</span>
|
||||
<span class="lineno">64</span> <span class="n">state_dict</span> <span class="o">=</span> <span class="n">hf_model</span><span class="o">.</span><span class="n">state_dict</span><span class="p">()</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">56</span> <span class="n">model</span><span class="p">:</span> <span class="n">GPTModel</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-9'>
|
||||
@ -201,18 +197,11 @@
|
||||
<div class='section-link'>
|
||||
<a href='#section-9'>#</a>
|
||||
</div>
|
||||
<p>Transformer embedding and prediction layer parameter mapping (<code class="highlight"><span></span><span class="n">hf</span><span class="p">:</span> <span class="n">ours</span></code>
|
||||
) </p>
|
||||
<p>Optimizer </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">67</span> <span class="n">mapping</span> <span class="o">=</span> <span class="p">{</span>
|
||||
<span class="lineno">68</span> <span class="s1">'transformer.wte.weight'</span><span class="p">:</span> <span class="s1">'token_embedding.weight'</span><span class="p">,</span>
|
||||
<span class="lineno">69</span> <span class="s1">'transformer.wpe.weight'</span><span class="p">:</span> <span class="s1">'position_embedding.weight'</span><span class="p">,</span>
|
||||
<span class="lineno">70</span> <span class="s1">'transformer.ln_f.weight'</span><span class="p">:</span> <span class="s1">'final_norm.weight'</span><span class="p">,</span>
|
||||
<span class="lineno">71</span> <span class="s1">'transformer.ln_f.bias'</span><span class="p">:</span> <span class="s1">'final_norm.bias'</span><span class="p">,</span>
|
||||
<span class="lineno">72</span> <span class="s1">'lm_head.weight'</span><span class="p">:</span> <span class="s1">'lm_head.weight'</span>
|
||||
<span class="lineno">73</span> <span class="p">}</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">58</span> <span class="n">optimizer</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">optim</span><span class="o">.</span><span class="n">Adam</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-10'>
|
||||
@ -220,24 +209,11 @@
|
||||
<div class='section-link'>
|
||||
<a href='#section-10'>#</a>
|
||||
</div>
|
||||
<p>Mapping (<code class="highlight"><span></span><span class="n">hf</span><span class="p">:</span> <span class="n">ours</span></code>
|
||||
) of decoder layers </p>
|
||||
<p>Cross entropy loss </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="mi">12</span><span class="p">):</span>
|
||||
<span class="lineno">77</span> <span class="n">mapping</span><span class="p">[</span><span class="sa">f</span><span class="s1">'transformer.h.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.ln_1.weight'</span><span class="p">]</span> <span class="o">=</span> <span class="sa">f</span><span class="s1">'blocks.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.pre_norm.weight'</span>
|
||||
<span class="lineno">78</span> <span class="n">mapping</span><span class="p">[</span><span class="sa">f</span><span class="s1">'transformer.h.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.ln_1.bias'</span><span class="p">]</span> <span class="o">=</span> <span class="sa">f</span><span class="s1">'blocks.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.pre_norm.bias'</span>
|
||||
<span class="lineno">79</span> <span class="n">mapping</span><span class="p">[</span><span class="sa">f</span><span class="s1">'transformer.h.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.attn.c_attn.weight'</span><span class="p">]</span> <span class="o">=</span> <span class="sa">f</span><span class="s1">'blocks.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.attn.c_att.weight'</span>
|
||||
<span class="lineno">80</span> <span class="n">mapping</span><span class="p">[</span><span class="sa">f</span><span class="s1">'transformer.h.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.attn.c_attn.bias'</span><span class="p">]</span> <span class="o">=</span> <span class="sa">f</span><span class="s1">'blocks.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.attn.c_att.bias'</span>
|
||||
<span class="lineno">81</span> <span class="n">mapping</span><span class="p">[</span><span class="sa">f</span><span class="s1">'transformer.h.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.attn.c_proj.weight'</span><span class="p">]</span> <span class="o">=</span> <span class="sa">f</span><span class="s1">'blocks.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.attn.c_proj.weight'</span>
|
||||
<span class="lineno">82</span> <span class="n">mapping</span><span class="p">[</span><span class="sa">f</span><span class="s1">'transformer.h.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.attn.c_proj.bias'</span><span class="p">]</span> <span class="o">=</span> <span class="sa">f</span><span class="s1">'blocks.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.attn.c_proj.bias'</span>
|
||||
<span class="lineno">83</span> <span class="n">mapping</span><span class="p">[</span><span class="sa">f</span><span class="s1">'transformer.h.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.ln_2.weight'</span><span class="p">]</span> <span class="o">=</span> <span class="sa">f</span><span class="s1">'blocks.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.post_norm.weight'</span>
|
||||
<span class="lineno">84</span> <span class="n">mapping</span><span class="p">[</span><span class="sa">f</span><span class="s1">'transformer.h.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.ln_2.bias'</span><span class="p">]</span> <span class="o">=</span> <span class="sa">f</span><span class="s1">'blocks.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.post_norm.bias'</span>
|
||||
<span class="lineno">85</span> <span class="n">mapping</span><span class="p">[</span><span class="sa">f</span><span class="s1">'transformer.h.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.mlp.c_fc.weight'</span><span class="p">]</span> <span class="o">=</span> <span class="sa">f</span><span class="s1">'blocks.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.ffn.c_fc.weight'</span>
|
||||
<span class="lineno">86</span> <span class="n">mapping</span><span class="p">[</span><span class="sa">f</span><span class="s1">'transformer.h.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.mlp.c_fc.bias'</span><span class="p">]</span> <span class="o">=</span> <span class="sa">f</span><span class="s1">'blocks.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.ffn.c_fc.bias'</span>
|
||||
<span class="lineno">87</span> <span class="n">mapping</span><span class="p">[</span><span class="sa">f</span><span class="s1">'transformer.h.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.mlp.c_proj.weight'</span><span class="p">]</span> <span class="o">=</span> <span class="sa">f</span><span class="s1">'blocks.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.ffn.c_proj.weight'</span>
|
||||
<span class="lineno">88</span> <span class="n">mapping</span><span class="p">[</span><span class="sa">f</span><span class="s1">'transformer.h.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.mlp.c_proj.bias'</span><span class="p">]</span> <span class="o">=</span> <span class="sa">f</span><span class="s1">'blocks.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.ffn.c_proj.bias'</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">60</span> <span class="n">loss_func</span> <span class="o">=</span> <span class="n">torch</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-11'>
|
||||
@ -245,32 +221,23 @@
|
||||
<div class='section-link'>
|
||||
<a href='#section-11'>#</a>
|
||||
</div>
|
||||
<p>Move the parameters based on mapping </p>
|
||||
<p>Dataloader </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">91</span> <span class="n">new_state_dict</span> <span class="o">=</span> <span class="p">{}</span>
|
||||
<span class="lineno">92</span> <span class="k">for</span> <span class="n">old_key</span><span class="p">,</span> <span class="n">new_key</span> <span class="ow">in</span> <span class="n">mapping</span><span class="o">.</span><span class="n">items</span><span class="p">():</span>
|
||||
<span class="lineno">93</span> <span class="k">if</span> <span class="n">old_key</span> <span class="ow">in</span> <span class="n">state_dict</span><span class="p">:</span>
|
||||
<span class="lineno">94</span> <span class="n">new_state_dict</span><span class="p">[</span><span class="n">new_key</span><span class="p">]</span> <span class="o">=</span> <span class="n">state_dict</span><span class="p">[</span><span class="n">old_key</span><span class="p">]</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">62</span> <span class="n">data_loader</span><span class="p">:</span> <span class="n">DataLoader</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-12'>
|
||||
<div class='docs'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-12'>#</a>
|
||||
</div>
|
||||
<p>GPT-2 hugging face uses 1D Convolution layers. We need to transpose those weights since we use linear layers </p>
|
||||
<h3>Load pre-trained <a href="https://huggingface.co/openai-community/gpt2">GPT-2 from huggingface</a></h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">97</span> <span class="n">convo_layers</span> <span class="o">=</span> <span class="p">([</span><span class="sa">f</span><span class="s1">'blocks.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.ffn.c_fc.weight'</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">12</span><span class="p">)]</span> <span class="o">+</span>
|
||||
<span class="lineno">98</span> <span class="p">[</span><span class="sa">f</span><span class="s1">'blocks.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.ffn.c_proj.weight'</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">12</span><span class="p">)]</span> <span class="o">+</span>
|
||||
<span class="lineno">99</span> <span class="p">[</span><span class="sa">f</span><span class="s1">'blocks.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.attn.c_att.weight'</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">12</span><span class="p">)]</span> <span class="o">+</span>
|
||||
<span class="lineno">100</span> <span class="p">[</span><span class="sa">f</span><span class="s1">'blocks.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.attn.c_proj.weight'</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">12</span><span class="p">)])</span>
|
||||
<span class="lineno">101</span>
|
||||
<span class="lineno">102</span> <span class="k">for</span> <span class="n">layer</span> <span class="ow">in</span> <span class="n">convo_layers</span><span class="p">:</span>
|
||||
<span class="lineno">103</span> <span class="n">new_state_dict</span><span class="p">[</span><span class="n">layer</span><span class="p">]</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">transpose</span><span class="p">(</span><span class="n">new_state_dict</span><span class="p">[</span><span class="n">layer</span><span class="p">],</span> <span class="mi">0</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">64</span> <span class="k">def</span> <span class="nf">_load_pretrained_weights</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-13'>
|
||||
@ -278,23 +245,31 @@
|
||||
<div class='section-link'>
|
||||
<a href='#section-13'>#</a>
|
||||
</div>
|
||||
<p>Load out model </p>
|
||||
<p>Load the huggingface model and get the parameters </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">106</span> <span class="bp">self</span><span class="o">.</span><span class="n">model</span><span class="o">.</span><span class="n">load_state_dict</span><span class="p">(</span><span class="n">new_state_dict</span><span class="p">,</span> <span class="n">strict</span><span class="o">=</span><span class="kc">False</span><span class="p">)</span> <span class="c1"># state dict does not have lora weights</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">70</span> <span class="n">hf_model</span> <span class="o">=</span> <span class="n">AutoModelForCausalLM</span><span class="o">.</span><span class="n">from_pretrained</span><span class="p">(</span><span class="s2">"gpt2"</span><span class="p">)</span>
|
||||
<span class="lineno">71</span> <span class="n">state_dict</span> <span class="o">=</span> <span class="n">hf_model</span><span class="o">.</span><span class="n">state_dict</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-14'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-14'>#</a>
|
||||
</div>
|
||||
<h3>Initialize the model, optimizer and dataloader</h3>
|
||||
<p>Transformer embedding and prediction layer parameter mapping (<code class="highlight"><span></span><span class="n">hf</span><span class="p">:</span> <span class="n">ours</span></code>
|
||||
) </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">108</span> <span class="k">def</span> <span class="nf">initialize</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">74</span> <span class="n">mapping</span> <span class="o">=</span> <span class="p">{</span>
|
||||
<span class="lineno">75</span> <span class="s1">'transformer.wte.weight'</span><span class="p">:</span> <span class="s1">'token_embedding.weight'</span><span class="p">,</span>
|
||||
<span class="lineno">76</span> <span class="s1">'transformer.wpe.weight'</span><span class="p">:</span> <span class="s1">'position_embedding.weight'</span><span class="p">,</span>
|
||||
<span class="lineno">77</span> <span class="s1">'transformer.ln_f.weight'</span><span class="p">:</span> <span class="s1">'final_norm.weight'</span><span class="p">,</span>
|
||||
<span class="lineno">78</span> <span class="s1">'transformer.ln_f.bias'</span><span class="p">:</span> <span class="s1">'final_norm.bias'</span><span class="p">,</span>
|
||||
<span class="lineno">79</span> <span class="s1">'lm_head.weight'</span><span class="p">:</span> <span class="s1">'lm_head.weight'</span>
|
||||
<span class="lineno">80</span> <span class="p">}</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-15'>
|
||||
@ -302,19 +277,24 @@
|
||||
<div class='section-link'>
|
||||
<a href='#section-15'>#</a>
|
||||
</div>
|
||||
<p>Initialize the model </p>
|
||||
<p>Mapping (<code class="highlight"><span></span><span class="n">hf</span><span class="p">:</span> <span class="n">ours</span></code>
|
||||
) of decoder layers </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">113</span> <span class="bp">self</span><span class="o">.</span><span class="n">model</span> <span class="o">=</span> <span class="n">GPTModel</span><span class="p">(</span>
|
||||
<span class="lineno">114</span> <span class="n">layer_norm_epsilon</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">layer_norm_epsilon</span><span class="p">,</span>
|
||||
<span class="lineno">115</span> <span class="n">n_embd</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">n_embed</span><span class="p">,</span>
|
||||
<span class="lineno">116</span> <span class="n">n_layer</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">n_layer</span><span class="p">,</span>
|
||||
<span class="lineno">117</span> <span class="n">n_positions</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">n_positions</span><span class="p">,</span>
|
||||
<span class="lineno">118</span> <span class="n">vocab_size</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">vocab_size</span><span class="p">,</span>
|
||||
<span class="lineno">119</span> <span class="n">r</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">lora_r</span><span class="p">,</span>
|
||||
<span class="lineno">120</span> <span class="p">)</span>
|
||||
<span class="lineno">121</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="bp">self</span><span class="o">.</span><span class="n">device</span><span class="p">)</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">83</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">12</span><span class="p">):</span>
|
||||
<span class="lineno">84</span> <span class="n">mapping</span><span class="p">[</span><span class="sa">f</span><span class="s1">'transformer.h.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.ln_1.weight'</span><span class="p">]</span> <span class="o">=</span> <span class="sa">f</span><span class="s1">'blocks.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.attn_norm.weight'</span>
|
||||
<span class="lineno">85</span> <span class="n">mapping</span><span class="p">[</span><span class="sa">f</span><span class="s1">'transformer.h.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.ln_1.bias'</span><span class="p">]</span> <span class="o">=</span> <span class="sa">f</span><span class="s1">'blocks.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.attn_norm.bias'</span>
|
||||
<span class="lineno">86</span> <span class="n">mapping</span><span class="p">[</span><span class="sa">f</span><span class="s1">'transformer.h.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.attn.c_attn.weight'</span><span class="p">]</span> <span class="o">=</span> <span class="sa">f</span><span class="s1">'blocks.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.attn.qkv_projection.weight'</span>
|
||||
<span class="lineno">87</span> <span class="n">mapping</span><span class="p">[</span><span class="sa">f</span><span class="s1">'transformer.h.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.attn.c_attn.bias'</span><span class="p">]</span> <span class="o">=</span> <span class="sa">f</span><span class="s1">'blocks.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.attn.qkv_projection.bias'</span>
|
||||
<span class="lineno">88</span> <span class="n">mapping</span><span class="p">[</span><span class="sa">f</span><span class="s1">'transformer.h.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.attn.c_proj.weight'</span><span class="p">]</span> <span class="o">=</span> <span class="sa">f</span><span class="s1">'blocks.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.attn.output_projection.weight'</span>
|
||||
<span class="lineno">89</span> <span class="n">mapping</span><span class="p">[</span><span class="sa">f</span><span class="s1">'transformer.h.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.attn.c_proj.bias'</span><span class="p">]</span> <span class="o">=</span> <span class="sa">f</span><span class="s1">'blocks.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.attn.output_projection.bias'</span>
|
||||
<span class="lineno">90</span> <span class="n">mapping</span><span class="p">[</span><span class="sa">f</span><span class="s1">'transformer.h.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.ln_2.weight'</span><span class="p">]</span> <span class="o">=</span> <span class="sa">f</span><span class="s1">'blocks.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.ffn_norm.weight'</span>
|
||||
<span class="lineno">91</span> <span class="n">mapping</span><span class="p">[</span><span class="sa">f</span><span class="s1">'transformer.h.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.ln_2.bias'</span><span class="p">]</span> <span class="o">=</span> <span class="sa">f</span><span class="s1">'blocks.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.ffn_norm.bias'</span>
|
||||
<span class="lineno">92</span> <span class="n">mapping</span><span class="p">[</span><span class="sa">f</span><span class="s1">'transformer.h.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.mlp.c_fc.weight'</span><span class="p">]</span> <span class="o">=</span> <span class="sa">f</span><span class="s1">'blocks.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.ffn.linear_in.weight'</span>
|
||||
<span class="lineno">93</span> <span class="n">mapping</span><span class="p">[</span><span class="sa">f</span><span class="s1">'transformer.h.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.mlp.c_fc.bias'</span><span class="p">]</span> <span class="o">=</span> <span class="sa">f</span><span class="s1">'blocks.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.ffn.linear_in.bias'</span>
|
||||
<span class="lineno">94</span> <span class="n">mapping</span><span class="p">[</span><span class="sa">f</span><span class="s1">'transformer.h.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.mlp.c_proj.weight'</span><span class="p">]</span> <span class="o">=</span> <span class="sa">f</span><span class="s1">'blocks.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.ffn.linear_out.weight'</span>
|
||||
<span class="lineno">95</span> <span class="n">mapping</span><span class="p">[</span><span class="sa">f</span><span class="s1">'transformer.h.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.mlp.c_proj.bias'</span><span class="p">]</span> <span class="o">=</span> <span class="sa">f</span><span class="s1">'blocks.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.ffn.linear_out.bias'</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-16'>
|
||||
@ -322,11 +302,14 @@
|
||||
<div class='section-link'>
|
||||
<a href='#section-16'>#</a>
|
||||
</div>
|
||||
<p>Load pre-trained model weights </p>
|
||||
<p>Move the parameters based on mapping </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">123</span> <span class="bp">self</span><span class="o">.</span><span class="n">_load_pretrained_weights</span><span class="p">()</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">98</span> <span class="n">new_state_dict</span> <span class="o">=</span> <span class="p">{}</span>
|
||||
<span class="lineno">99</span> <span class="k">for</span> <span class="n">old_key</span><span class="p">,</span> <span class="n">new_key</span> <span class="ow">in</span> <span class="n">mapping</span><span class="o">.</span><span class="n">items</span><span class="p">():</span>
|
||||
<span class="lineno">100</span> <span class="k">if</span> <span class="n">old_key</span> <span class="ow">in</span> <span class="n">state_dict</span><span class="p">:</span>
|
||||
<span class="lineno">101</span> <span class="n">new_state_dict</span><span class="p">[</span><span class="n">new_key</span><span class="p">]</span> <span class="o">=</span> <span class="n">state_dict</span><span class="p">[</span><span class="n">old_key</span><span class="p">]</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-17'>
|
||||
@ -334,11 +317,17 @@
|
||||
<div class='section-link'>
|
||||
<a href='#section-17'>#</a>
|
||||
</div>
|
||||
<p>Initialize the optimizer </p>
|
||||
<p>GPT-2 hugging face uses 1D Convolution layers. We need to transpose those weights since we use linear layers </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">126</span> <span class="bp">self</span><span class="o">.</span><span class="n">optimizer</span> <span class="o">=</span> <span class="n">Adam</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">model</span><span class="o">.</span><span class="n">parameters</span><span class="p">(),</span> <span class="n">lr</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">learning_rate</span><span class="p">)</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">104</span> <span class="n">convo_layers</span> <span class="o">=</span> <span class="p">([</span><span class="sa">f</span><span class="s1">'blocks.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.ffn.linear_in.weight'</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">12</span><span class="p">)]</span> <span class="o">+</span>
|
||||
<span class="lineno">105</span> <span class="p">[</span><span class="sa">f</span><span class="s1">'blocks.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.ffn.linear_out.weight'</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">12</span><span class="p">)]</span> <span class="o">+</span>
|
||||
<span class="lineno">106</span> <span class="p">[</span><span class="sa">f</span><span class="s1">'blocks.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.attn.qkv_projection.weight'</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">12</span><span class="p">)]</span> <span class="o">+</span>
|
||||
<span class="lineno">107</span> <span class="p">[</span><span class="sa">f</span><span class="s1">'blocks.</span><span class="si">{</span><span class="n">i</span><span class="si">}</span><span class="s1">.attn.output_projection.weight'</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">12</span><span class="p">)])</span>
|
||||
<span class="lineno">108</span>
|
||||
<span class="lineno">109</span> <span class="k">for</span> <span class="n">layer</span> <span class="ow">in</span> <span class="n">convo_layers</span><span class="p">:</span>
|
||||
<span class="lineno">110</span> <span class="n">new_state_dict</span><span class="p">[</span><span class="n">layer</span><span class="p">]</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">transpose</span><span class="p">(</span><span class="n">new_state_dict</span><span class="p">[</span><span class="n">layer</span><span class="p">],</span> <span class="mi">0</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-18'>
|
||||
@ -346,69 +335,58 @@
|
||||
<div class='section-link'>
|
||||
<a href='#section-18'>#</a>
|
||||
</div>
|
||||
<p>Initialize the data loader </p>
|
||||
<p>Load out model. We use <code class="highlight"><span></span><span class="n">strict</span> <span class="o">=</span> <span class="kc">False</span></code>
|
||||
because the state does not have LoRA weights </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">129</span> <span class="bp">self</span><span class="o">.</span><span class="n">data_loader</span> <span class="o">=</span> <span class="n">DataLoader</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">text</span><span class="p">,</span> <span class="n">batch_size</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">batch_size</span><span class="p">,</span> <span class="n">shuffle</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">113</span> <span class="n">missing_keys</span><span class="p">,</span> <span class="n">unexpected_keys</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">model</span><span class="o">.</span><span class="n">load_state_dict</span><span class="p">(</span><span class="n">new_state_dict</span><span class="p">,</span> <span class="n">strict</span><span class="o">=</span><span class="kc">False</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-19'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-19'>#</a>
|
||||
</div>
|
||||
<h3>Training loop</h3>
|
||||
<p>make sure that only lora weights are not loaded </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">131</span> <span class="k">def</span> <span class="nf">run</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">116</span> <span class="k">assert</span> <span class="nb">all</span><span class="p">(</span><span class="s1">'lora'</span> <span class="ow">in</span> <span class="n">key</span> <span class="k">for</span> <span class="n">key</span> <span class="ow">in</span> <span class="n">missing_keys</span><span class="p">)</span>
|
||||
<span class="lineno">117</span> <span class="k">assert</span> <span class="ow">not</span> <span class="n">unexpected_keys</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-20'>
|
||||
<div class='docs'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-20'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">136</span> <span class="k">for</span> <span class="n">_</span> <span class="ow">in</span> <span class="n">monit</span><span class="o">.</span><span class="n">loop</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">epochs</span><span class="p">):</span>
|
||||
<span class="lineno">137</span> <span class="k">for</span> <span class="n">i</span><span class="p">,</span> <span class="n">batch</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">data_loader</span><span class="p">):</span>
|
||||
<span class="lineno">138</span> <span class="n">inputs</span> <span class="o">=</span> <span class="n">batch</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span>
|
||||
<span class="lineno">139</span> <span class="n">inputs</span> <span class="o">=</span> <span class="n">inputs</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">140</span> <span class="n">labels</span> <span class="o">=</span> <span class="n">inputs</span><span class="o">.</span><span class="n">clone</span><span class="p">()</span>
|
||||
<span class="lineno">141</span>
|
||||
<span class="lineno">142</span> <span class="n">outputs</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">inputs</span><span class="p">)</span>
|
||||
<span class="lineno">143</span>
|
||||
<span class="lineno">144</span> <span class="n">shift_logits</span> <span class="o">=</span> <span class="n">outputs</span><span class="p">[</span><span class="o">...</span><span class="p">,</span> <span class="p">:</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="p">:]</span>
|
||||
<span class="lineno">145</span> <span class="n">shift_labels</span> <span class="o">=</span> <span class="n">labels</span><span class="p">[</span><span class="o">...</span><span class="p">,</span> <span class="mi">1</span><span class="p">:]</span>
|
||||
<span class="lineno">146</span>
|
||||
<span class="lineno">147</span> <span class="n">loss</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">criterion</span><span class="p">(</span><span class="n">shift_logits</span><span class="o">.</span><span class="n">reshape</span><span class="p">(</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="n">shift_logits</span><span class="o">.</span><span class="n">size</span><span class="p">(</span><span class="o">-</span><span class="mi">1</span><span class="p">)),</span> <span class="n">shift_labels</span><span class="o">.</span><span class="n">reshape</span><span class="p">(</span><span class="o">-</span><span class="mi">1</span><span class="p">))</span>
|
||||
<span class="lineno">148</span>
|
||||
<span class="lineno">149</span> <span class="bp">self</span><span class="o">.</span><span class="n">optimizer</span><span class="o">.</span><span class="n">zero_grad</span><span class="p">()</span>
|
||||
<span class="lineno">150</span> <span class="n">loss</span><span class="o">.</span><span class="n">backward</span><span class="p">()</span>
|
||||
<span class="lineno">151</span> <span class="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>
|
||||
<span class="lineno">152</span>
|
||||
<span class="lineno">153</span> <span class="n">tracker</span><span class="o">.</span><span class="n">add</span><span class="p">({</span><span class="s1">'loss'</span><span class="p">:</span> <span class="n">loss</span><span class="p">})</span>
|
||||
<span class="lineno">154</span>
|
||||
<span class="lineno">155</span> <span class="n">tracker</span><span class="o">.</span><span class="n">save</span><span class="p">()</span>
|
||||
<span class="lineno">156</span> <span class="n">tracker</span><span class="o">.</span><span class="n">add_global_step</span><span class="p">()</span>
|
||||
<span class="lineno">157</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-21'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-21'>#</a>
|
||||
</div>
|
||||
<h3>Tiny Shakespeare dataset</h3>
|
||||
<p>It will download from the url if not present</p>
|
||||
<h3>Initialize the model, optimizer and dataloader</h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">160</span><span class="nd">@option</span><span class="p">(</span><span class="n">Trainer</span><span class="o">.</span><span class="n">text</span><span class="p">)</span>
|
||||
<span class="lineno">161</span><span class="k">def</span> <span class="nf">tiny_shakespeare</span><span class="p">(</span><span class="n">c</span><span class="p">:</span> <span class="n">Trainer</span><span class="p">):</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">119</span> <span class="k">def</span> <span class="nf">initialize</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-21'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-21'>#</a>
|
||||
</div>
|
||||
<p>Initialize the <a href="gpt2.html">GPT2 model</a> </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">124</span> <span class="bp">self</span><span class="o">.</span><span class="n">model</span> <span class="o">=</span> <span class="n">GPTModel</span><span class="p">(</span>
|
||||
<span class="lineno">125</span> <span class="n">layer_norm_epsilon</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">layer_norm_epsilon</span><span class="p">,</span>
|
||||
<span class="lineno">126</span> <span class="n">d_model</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">d_model</span><span class="p">,</span>
|
||||
<span class="lineno">127</span> <span class="n">n_layers</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">n_layers</span><span class="p">,</span>
|
||||
<span class="lineno">128</span> <span class="n">n_heads</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">n_heads</span><span class="p">,</span>
|
||||
<span class="lineno">129</span> <span class="n">n_positions</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">n_positions</span><span class="p">,</span>
|
||||
<span class="lineno">130</span> <span class="n">vocab_size</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">vocab_size</span><span class="p">,</span>
|
||||
<span class="lineno">131</span> <span class="n">r</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">lora_r</span><span class="p">,</span>
|
||||
<span class="lineno">132</span> <span class="p">)</span>
|
||||
<span class="lineno">133</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="bp">self</span><span class="o">.</span><span class="n">device</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-22'>
|
||||
@ -416,20 +394,205 @@
|
||||
<div class='section-link'>
|
||||
<a href='#section-22'>#</a>
|
||||
</div>
|
||||
<p>Load pre-trained model weights </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">135</span> <span class="bp">self</span><span class="o">.</span><span class="n">_load_pretrained_weights</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-23'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-23'>#</a>
|
||||
</div>
|
||||
<p>Initialize the optimizer </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">138</span> <span class="bp">self</span><span class="o">.</span><span class="n">optimizer</span> <span class="o">=</span> <span class="n">Adam</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">model</span><span class="o">.</span><span class="n">parameters</span><span class="p">(),</span> <span class="n">lr</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">learning_rate</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-24'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-24'>#</a>
|
||||
</div>
|
||||
<p>Initialize the data loader </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">141</span> <span class="bp">self</span><span class="o">.</span><span class="n">data_loader</span> <span class="o">=</span> <span class="n">DataLoader</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">text</span><span class="p">,</span> <span class="n">batch_size</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">batch_size</span><span class="p">,</span> <span class="n">shuffle</span><span class="o">=</span><span class="kc">True</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>
|
||||
<h3>Training loop</h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">143</span> <span class="k">def</span> <span class="nf">run</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-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">167</span> <span class="n">path</span> <span class="o">=</span> <span class="n">lab</span><span class="o">.</span><span class="n">get_data_path</span><span class="p">()</span> <span class="o">/</span> <span class="s1">'tiny_shakespeare.txt'</span>
|
||||
<span class="lineno">168</span> <span class="k">if</span> <span class="ow">not</span> <span class="n">path</span><span class="o">.</span><span class="n">exists</span><span class="p">():</span>
|
||||
<span class="lineno">169</span> <span class="n">download_file</span><span class="p">(</span><span class="s2">"https://raw.githubusercontent.com/karpathy/char-rnn/master/data/tinyshakespeare/input.txt"</span><span class="p">,</span> <span class="n">path</span><span class="p">)</span>
|
||||
<span class="lineno">170</span> <span class="k">with</span> <span class="nb">open</span><span class="p">(</span><span class="n">path</span><span class="p">,</span> <span class="s1">'r'</span><span class="p">,</span> <span class="n">encoding</span><span class="o">=</span><span class="s1">'utf-8'</span><span class="p">)</span> <span class="k">as</span> <span class="n">f</span><span class="p">:</span>
|
||||
<span class="lineno">171</span> <span class="n">text</span> <span class="o">=</span> <span class="n">f</span><span class="o">.</span><span class="n">read</span><span class="p">()</span>
|
||||
<span class="lineno">172</span>
|
||||
<span class="lineno">173</span> <span class="n">tokens</span> <span class="o">=</span> <span class="n">c</span><span class="o">.</span><span class="n">tokenizer</span><span class="o">.</span><span class="n">encode</span><span class="p">(</span><span class="n">text</span><span class="p">)</span>
|
||||
<span class="lineno">174</span> <span class="n">num_batches</span> <span class="o">=</span> <span class="nb">len</span><span class="p">(</span><span class="n">tokens</span><span class="p">)</span> <span class="o">//</span> <span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">batch_size</span> <span class="o">*</span> <span class="n">c</span><span class="o">.</span><span class="n">context_len</span><span class="p">)</span>
|
||||
<span class="lineno">175</span> <span class="n">tokens</span> <span class="o">=</span> <span class="n">tokens</span><span class="p">[:</span><span class="n">num_batches</span> <span class="o">*</span> <span class="n">c</span><span class="o">.</span><span class="n">batch_size</span> <span class="o">*</span> <span class="n">c</span><span class="o">.</span><span class="n">context_len</span><span class="p">]</span>
|
||||
<span class="lineno">176</span> <span class="n">input_ids</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">tensor</span><span class="p">(</span><span class="n">tokens</span><span class="p">)</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">c</span><span class="o">.</span><span class="n">context_len</span><span class="p">)</span>
|
||||
<span class="lineno">177</span> <span class="k">return</span> <span class="n">TensorDataset</span><span class="p">(</span><span class="n">input_ids</span><span class="p">)</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">148</span> <span class="k">for</span> <span class="n">_</span> <span class="ow">in</span> <span class="n">monit</span><span class="o">.</span><span class="n">loop</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">epochs</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-27'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-27'>#</a>
|
||||
</div>
|
||||
<p><code class="highlight"><span></span><span class="n">inputs</span></code>
|
||||
has shape <code class="highlight"><span></span><span class="p">[</span><span class="n">batch_size</span><span class="p">,</span> <span class="n">seq_len</span><span class="p">]</span></code>
|
||||
</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">150</span> <span class="k">for</span> <span class="p">(</span><span class="n">inputs</span><span class="p">,)</span> <span class="ow">in</span> <span class="n">monit</span><span class="o">.</span><span class="n">iterate</span><span class="p">(</span><span class="s1">'Train'</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">data_loader</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-28'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-28'>#</a>
|
||||
</div>
|
||||
<p>Move <code class="highlight"><span></span><span class="n">inputs</span></code>
|
||||
to device </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">152</span> <span class="n">inputs</span> <span class="o">=</span> <span class="n">inputs</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>Call the model, with the all but the last token </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">154</span> <span class="n">logits</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">inputs</span><span class="p">[:,</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-30'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-30'>#</a>
|
||||
</div>
|
||||
<p>Get cross entropy loss </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">156</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">logits</span><span class="o">.</span><span class="n">reshape</span><span class="p">(</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="n">logits</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">inputs</span><span class="p">[:,</span> <span class="mi">1</span><span class="p">:]</span><span class="o">.</span><span class="n">reshape</span><span class="p">(</span><span class="o">-</span><span class="mi">1</span><span class="p">))</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-31'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-31'>#</a>
|
||||
</div>
|
||||
<p>Make gradients 0 </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">159</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>Compute gradients </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">161</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>Optimize </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">163</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>Log the loss </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">166</span> <span class="n">tracker</span><span class="o">.</span><span class="n">save</span><span class="p">({</span><span class="s1">'loss'</span><span class="p">:</span> <span class="n">loss</span><span class="p">})</span>
|
||||
<span class="lineno">167</span> <span class="n">tracker</span><span class="o">.</span><span class="n">add_global_step</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-35'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-35'>#</a>
|
||||
</div>
|
||||
<p> </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">169</span> <span class="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-36'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-36'>#</a>
|
||||
</div>
|
||||
<h3>Tiny Shakespeare dataset</h3>
|
||||
<p>It will download from the url if not present</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">172</span><span class="nd">@option</span><span class="p">(</span><span class="n">Trainer</span><span class="o">.</span><span class="n">text</span><span class="p">)</span>
|
||||
<span class="lineno">173</span><span class="k">def</span> <span class="nf">tiny_shakespeare</span><span class="p">(</span><span class="n">c</span><span class="p">:</span> <span class="n">Trainer</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>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">179</span> <span class="n">path</span> <span class="o">=</span> <span class="n">lab</span><span class="o">.</span><span class="n">get_data_path</span><span class="p">()</span> <span class="o">/</span> <span class="s1">'tiny_shakespeare.txt'</span>
|
||||
<span class="lineno">180</span> <span class="k">if</span> <span class="ow">not</span> <span class="n">path</span><span class="o">.</span><span class="n">exists</span><span class="p">():</span>
|
||||
<span class="lineno">181</span> <span class="n">download_file</span><span class="p">(</span><span class="s2">"https://raw.githubusercontent.com/karpathy/char-rnn/master/data/tinyshakespeare/input.txt"</span><span class="p">,</span> <span class="n">path</span><span class="p">)</span>
|
||||
<span class="lineno">182</span> <span class="k">with</span> <span class="nb">open</span><span class="p">(</span><span class="n">path</span><span class="p">,</span> <span class="s1">'r'</span><span class="p">,</span> <span class="n">encoding</span><span class="o">=</span><span class="s1">'utf-8'</span><span class="p">)</span> <span class="k">as</span> <span class="n">f</span><span class="p">:</span>
|
||||
<span class="lineno">183</span> <span class="n">text</span> <span class="o">=</span> <span class="n">f</span><span class="o">.</span><span class="n">read</span><span class="p">()</span>
|
||||
<span class="lineno">184</span>
|
||||
<span class="lineno">185</span> <span class="n">tokens</span> <span class="o">=</span> <span class="n">c</span><span class="o">.</span><span class="n">tokenizer</span><span class="o">.</span><span class="n">encode</span><span class="p">(</span><span class="n">text</span><span class="p">)</span>
|
||||
<span class="lineno">186</span> <span class="n">num_batches</span> <span class="o">=</span> <span class="nb">len</span><span class="p">(</span><span class="n">tokens</span><span class="p">)</span> <span class="o">//</span> <span class="p">(</span><span class="n">c</span><span class="o">.</span><span class="n">batch_size</span> <span class="o">*</span> <span class="n">c</span><span class="o">.</span><span class="n">context_len</span><span class="p">)</span>
|
||||
<span class="lineno">187</span> <span class="n">tokens</span> <span class="o">=</span> <span class="n">tokens</span><span class="p">[:</span><span class="n">num_batches</span> <span class="o">*</span> <span class="n">c</span><span class="o">.</span><span class="n">batch_size</span> <span class="o">*</span> <span class="n">c</span><span class="o">.</span><span class="n">context_len</span><span class="p">]</span>
|
||||
<span class="lineno">188</span> <span class="n">input_ids</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">tensor</span><span class="p">(</span><span class="n">tokens</span><span class="p">)</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">c</span><span class="o">.</span><span class="n">context_len</span><span class="p">)</span>
|
||||
<span class="lineno">189</span> <span class="k">return</span> <span class="n">TensorDataset</span><span class="p">(</span><span class="n">input_ids</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
|
||||
@ -3,24 +3,24 @@
|
||||
<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="description" content="GPT-2 implementation with LoRA modules"/>
|
||||
|
||||
<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="gpt2.py"/>
|
||||
<meta name="twitter:description" content=""/>
|
||||
<meta name="twitter:title" content="GPT-2 with LoRA"/>
|
||||
<meta name="twitter:description" content="GPT-2 implementation with LoRA modules"/>
|
||||
<meta name="twitter:site" content="@labmlai"/>
|
||||
<meta name="twitter:creator" content="@labmlai"/>
|
||||
|
||||
<meta property="og:url" content="https://nn.labml.ai/lora/gpt2.html"/>
|
||||
<meta property="og:title" content="gpt2.py"/>
|
||||
<meta property="og:title" content="GPT-2 with LoRA"/>
|
||||
<meta property="og:image" content="https://avatars1.githubusercontent.com/u/64068543?s=400&v=4"/>
|
||||
<meta property="og:site_name" content="gpt2.py"/>
|
||||
<meta property="og:site_name" content="GPT-2 with LoRA"/>
|
||||
<meta property="og:type" content="object"/>
|
||||
<meta property="og:title" content="gpt2.py"/>
|
||||
<meta property="og:description" content=""/>
|
||||
<meta property="og:title" content="GPT-2 with LoRA"/>
|
||||
<meta property="og:description" content="GPT-2 implementation with LoRA modules"/>
|
||||
|
||||
<title>gpt2.py</title>
|
||||
<title>GPT-2 with LoRA</title>
|
||||
<link rel="shortcut icon" href="/icon.png"/>
|
||||
<link rel="stylesheet" href="../pylit.css?v=1">
|
||||
<link rel="canonical" href="https://nn.labml.ai/lora/gpt2.html"/>
|
||||
@ -66,39 +66,48 @@
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-0'>
|
||||
<div class='docs'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-0'>#</a>
|
||||
</div>
|
||||
|
||||
<h1>GPT-2 with <a href="index.html">LoRA modules</a></h1>
|
||||
<p>Here's <a href="experiment.html">the training code</a> for training a GPT2 model with LoRA on Tiny Shakespeare dataset.</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">1</span><span></span><span class="kn">import</span> <span class="nn">torch</span>
|
||||
<span class="lineno">2</span><span class="kn">import</span> <span class="nn">torch.nn</span> <span class="k">as</span> <span class="nn">nn</span>
|
||||
<span class="lineno">3</span><span class="kn">from</span> <span class="nn">labml_nn.lora</span> <span class="kn">import</span> <span class="n">Linear</span><span class="p">,</span> <span class="n">Embedding</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">13</span><span></span><span class="kn">import</span> <span class="nn">torch</span>
|
||||
<span class="lineno">14</span><span class="kn">import</span> <span class="nn">torch.nn</span> <span class="k">as</span> <span class="nn">nn</span>
|
||||
<span class="lineno">15</span>
|
||||
<span class="lineno">16</span><span class="kn">from</span> <span class="nn">labml_nn.lora</span> <span class="kn">import</span> <span class="n">Linear</span><span class="p">,</span> <span class="n">Embedding</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-1'>
|
||||
<div class='docs'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-1'>#</a>
|
||||
</div>
|
||||
|
||||
<h3>Feedforward Network</h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">6</span><span class="k">class</span> <span class="nc">FFN</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 class="highlight"><pre><span class="lineno">19</span><span class="k">class</span> <span class="nc">FFN</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='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-2'>#</a>
|
||||
</div>
|
||||
|
||||
<ul><li><code class="highlight"><span></span><span class="n">d_model</span></code>
|
||||
is the number of dimensions </li>
|
||||
<li><code class="highlight"><span></span><span class="n">d_ff</span></code>
|
||||
is the size of the hidden dimension </li>
|
||||
<li><code class="highlight"><span></span><span class="n">r</span></code>
|
||||
is the lora rank</li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">7</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">dim</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">n_embed</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">r</span><span class="p">:</span> <span class="nb">int</span><span class="p">):</span>
|
||||
<span class="lineno">8</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 class="highlight"><pre><span class="lineno">24</span> <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">d_model</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">d_ff</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">r</span><span class="p">:</span> <span class="nb">int</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-3'>
|
||||
@ -106,11 +115,10 @@
|
||||
<div class='section-link'>
|
||||
<a href='#section-3'>#</a>
|
||||
</div>
|
||||
<p>lin1 </p>
|
||||
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">10</span> <span class="bp">self</span><span class="o">.</span><span class="n">c_fc</span> <span class="o">=</span> <span class="n">Linear</span><span class="p">(</span><span class="n">n_embed</span><span class="p">,</span> <span class="n">dim</span><span class="p">,</span> <span class="n">r</span><span class="o">=</span><span class="n">r</span><span class="p">,</span> <span class="n">bias</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">30</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-4'>
|
||||
@ -118,27 +126,27 @@
|
||||
<div class='section-link'>
|
||||
<a href='#section-4'>#</a>
|
||||
</div>
|
||||
<p>lin2 </p>
|
||||
<p>The linear layers and the activation </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">12</span> <span class="bp">self</span><span class="o">.</span><span class="n">c_proj</span> <span class="o">=</span> <span class="n">Linear</span><span class="p">(</span><span class="n">dim</span><span class="p">,</span> <span class="n">n_embed</span><span class="p">,</span> <span class="n">r</span><span class="o">=</span><span class="n">r</span><span class="p">,</span> <span class="n">bias</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
|
||||
<span class="lineno">13</span> <span class="bp">self</span><span class="o">.</span><span class="n">act</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">functional</span><span class="o">.</span><span class="n">gelu</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">33</span> <span class="bp">self</span><span class="o">.</span><span class="n">linear_in</span> <span class="o">=</span> <span class="n">Linear</span><span class="p">(</span><span class="n">d_model</span><span class="p">,</span> <span class="n">d_ff</span><span class="p">,</span> <span class="n">r</span><span class="o">=</span><span class="n">r</span><span class="p">,</span> <span class="n">bias</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
|
||||
<span class="lineno">34</span> <span class="bp">self</span><span class="o">.</span><span class="n">linear_out</span> <span class="o">=</span> <span class="n">Linear</span><span class="p">(</span><span class="n">d_ff</span><span class="p">,</span> <span class="n">d_model</span><span class="p">,</span> <span class="n">r</span><span class="o">=</span><span class="n">r</span><span class="p">,</span> <span class="n">bias</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
|
||||
<span class="lineno">35</span> <span class="bp">self</span><span class="o">.</span><span class="n">act</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">GELU</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-5'>
|
||||
<div class='docs'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-5'>#</a>
|
||||
</div>
|
||||
|
||||
<ul><li><code class="highlight"><span></span><span class="n">x</span></code>
|
||||
is the embeddings tensor with shape <code class="highlight"><span></span><span class="p">[</span><span class="n">batch_size</span><span class="p">,</span> <span class="n">seq_len</span><span class="p">,</span> <span class="n">d_model</span><span class="p">]</span></code>
|
||||
</li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">15</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">hidden_states</span><span class="p">):</span>
|
||||
<span class="lineno">16</span> <span class="n">hidden_states</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">c_fc</span><span class="p">(</span><span class="n">hidden_states</span><span class="p">)</span>
|
||||
<span class="lineno">17</span> <span class="n">hidden_states</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">act</span><span class="p">(</span><span class="n">hidden_states</span><span class="p">)</span>
|
||||
<span class="lineno">18</span> <span class="n">hidden_states</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">c_proj</span><span class="p">(</span><span class="n">hidden_states</span><span class="p">)</span>
|
||||
<span class="lineno">19</span> <span class="k">return</span> <span class="n">hidden_states</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">37</span> <span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">)</span> <span class="o">-></span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">:</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-6'>
|
||||
@ -149,35 +157,39 @@
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">22</span><span class="k">class</span> <span class="nc">MultiHeadAttention</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 class="highlight"><pre><span class="lineno">41</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">linear_in</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
|
||||
<span class="lineno">42</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">act</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
|
||||
<span class="lineno">43</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">linear_out</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
|
||||
<span class="lineno">44</span> <span class="k">return</span> <span class="n">x</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-7'>
|
||||
<div class='docs'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-7'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">23</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_embed</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">r</span><span class="p">:</span> <span class="nb">int</span><span class="p">):</span>
|
||||
<span class="lineno">24</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
|
||||
<span class="lineno">25</span> <span class="bp">self</span><span class="o">.</span><span class="n">embed_dim</span> <span class="o">=</span> <span class="n">n_embed</span>
|
||||
<span class="lineno">26</span> <span class="bp">self</span><span class="o">.</span><span class="n">num_heads</span> <span class="o">=</span> <span class="n">n_embed</span>
|
||||
<span class="lineno">27</span> <span class="bp">self</span><span class="o">.</span><span class="n">head_dim</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">embed_dim</span> <span class="o">//</span> <span class="bp">self</span><span class="o">.</span><span class="n">num_heads</span>
|
||||
<span class="lineno">28</span> <span class="bp">self</span><span class="o">.</span><span class="n">split_size</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">embed_dim</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>qkv </p>
|
||||
<h3>Multi-Head Attention</h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">31</span> <span class="bp">self</span><span class="o">.</span><span class="n">c_att</span> <span class="o">=</span> <span class="n">Linear</span><span class="p">(</span><span class="n">n_embed</span><span class="p">,</span> <span class="n">n_embed</span> <span class="o">*</span> <span class="mi">3</span><span class="p">,</span> <span class="n">r</span><span class="o">=</span><span class="n">r</span><span class="p">,</span> <span class="n">bias</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">47</span><span class="k">class</span> <span class="nc">MultiHeadAttention</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-8'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-8'>#</a>
|
||||
</div>
|
||||
<ul><li><code class="highlight"><span></span><span class="n">d_model</span></code>
|
||||
is the number of dimensions in the embeddings </li>
|
||||
<li><code class="highlight"><span></span><span class="n">n_heads</span></code>
|
||||
is the number of heads </li>
|
||||
<li><code class="highlight"><span></span><span class="n">r</span></code>
|
||||
is the lora rank</li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">52</span> <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">d_model</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">n_heads</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">r</span><span class="p">:</span> <span class="nb">int</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-9'>
|
||||
@ -185,23 +197,25 @@
|
||||
<div class='section-link'>
|
||||
<a href='#section-9'>#</a>
|
||||
</div>
|
||||
<p>out </p>
|
||||
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">33</span> <span class="bp">self</span><span class="o">.</span><span class="n">c_proj</span> <span class="o">=</span> <span class="n">Linear</span><span class="p">(</span><span class="n">n_embed</span><span class="p">,</span> <span class="n">n_embed</span><span class="p">,</span> <span class="n">r</span><span class="o">=</span><span class="n">r</span><span class="p">,</span> <span class="n">bias</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">58</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
|
||||
<span class="lineno">59</span> <span class="bp">self</span><span class="o">.</span><span class="n">d_model</span> <span class="o">=</span> <span class="n">d_model</span>
|
||||
<span class="lineno">60</span> <span class="bp">self</span><span class="o">.</span><span class="n">n_heads</span> <span class="o">=</span> <span class="n">n_heads</span>
|
||||
<span class="lineno">61</span> <span class="bp">self</span><span class="o">.</span><span class="n">d_head</span> <span class="o">=</span> <span class="n">d_model</span> <span class="o">//</span> <span class="n">n_heads</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-10'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-10'>#</a>
|
||||
</div>
|
||||
<p> Splits hidden_size dim into attn_head_size and num_heads</p>
|
||||
<p>Linear transformation for QKV </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">35</span> <span class="k">def</span> <span class="nf">_split_heads</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">tensor</span><span class="p">,</span> <span class="n">num_heads</span><span class="p">,</span> <span class="n">attn_head_size</span><span class="p">):</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">64</span> <span class="bp">self</span><span class="o">.</span><span class="n">qkv_projection</span> <span class="o">=</span> <span class="n">Linear</span><span class="p">(</span><span class="n">d_model</span><span class="p">,</span> <span class="n">d_model</span> <span class="o">*</span> <span class="mi">3</span><span class="p">,</span> <span class="n">r</span><span class="o">=</span><span class="n">r</span><span class="p">,</span> <span class="n">bias</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-11'>
|
||||
@ -209,46 +223,25 @@
|
||||
<div class='section-link'>
|
||||
<a href='#section-11'>#</a>
|
||||
</div>
|
||||
|
||||
<p>Output projection </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">39</span> <span class="n">new_shape</span> <span class="o">=</span> <span class="n">tensor</span><span class="o">.</span><span class="n">size</span><span class="p">()[:</span><span class="o">-</span><span class="mi">1</span><span class="p">]</span> <span class="o">+</span> <span class="p">(</span><span class="n">num_heads</span><span class="p">,</span> <span class="n">attn_head_size</span><span class="p">)</span>
|
||||
<span class="lineno">40</span> <span class="n">tensor</span> <span class="o">=</span> <span class="n">tensor</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="n">new_shape</span><span class="p">)</span>
|
||||
<span class="lineno">41</span> <span class="k">return</span> <span class="n">tensor</span><span class="o">.</span><span class="n">permute</span><span class="p">(</span><span class="mi">0</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">3</span><span class="p">)</span> <span class="c1"># (batch, head, seq_length, head_features)</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">66</span> <span class="bp">self</span><span class="o">.</span><span class="n">output_projection</span> <span class="o">=</span> <span class="n">Linear</span><span class="p">(</span><span class="n">d_model</span><span class="p">,</span> <span class="n">d_model</span><span class="p">,</span> <span class="n">r</span><span class="o">=</span><span class="n">r</span><span class="p">,</span> <span class="n">bias</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-12'>
|
||||
<div class='docs'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-12'>#</a>
|
||||
</div>
|
||||
|
||||
<ul><li><code class="highlight"><span></span><span class="n">x</span></code>
|
||||
is the tensor with shape <code class="highlight"><span></span><span class="p">[</span><span class="n">batch_size</span><span class="p">,</span> <span class="n">seq_len</span><span class="p">,</span> <span class="n">d_model</span><span class="p">]</span></code>
|
||||
</li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">43</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">hidden_states</span><span class="p">):</span>
|
||||
<span class="lineno">44</span> <span class="n">batch_size</span><span class="p">,</span> <span class="n">seq_length</span><span class="p">,</span> <span class="n">_</span> <span class="o">=</span> <span class="n">hidden_states</span><span class="o">.</span><span class="n">size</span><span class="p">()</span>
|
||||
<span class="lineno">45</span>
|
||||
<span class="lineno">46</span> <span class="n">query</span><span class="p">,</span> <span class="n">key</span><span class="p">,</span> <span class="n">value</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">c_att</span><span class="p">(</span><span class="n">hidden_states</span><span class="p">)</span><span class="o">.</span><span class="n">split</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">split_size</span><span class="p">,</span> <span class="n">dim</span><span class="o">=</span><span class="mi">2</span><span class="p">)</span>
|
||||
<span class="lineno">47</span>
|
||||
<span class="lineno">48</span> <span class="n">query</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">_split_heads</span><span class="p">(</span><span class="n">query</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">num_heads</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">head_dim</span><span class="p">)</span>
|
||||
<span class="lineno">49</span> <span class="n">key</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">_split_heads</span><span class="p">(</span><span class="n">key</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">num_heads</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">head_dim</span><span class="p">)</span>
|
||||
<span class="lineno">50</span> <span class="n">value</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">_split_heads</span><span class="p">(</span><span class="n">value</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">num_heads</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">head_dim</span><span class="p">)</span>
|
||||
<span class="lineno">51</span>
|
||||
<span class="lineno">52</span> <span class="n">attn_output</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">nn</span><span class="o">.</span><span class="n">functional</span><span class="o">.</span><span class="n">scaled_dot_product_attention</span><span class="p">(</span>
|
||||
<span class="lineno">53</span> <span class="n">query</span><span class="p">,</span>
|
||||
<span class="lineno">54</span> <span class="n">key</span><span class="p">,</span>
|
||||
<span class="lineno">55</span> <span class="n">value</span><span class="p">,</span>
|
||||
<span class="lineno">56</span> <span class="n">attn_mask</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
|
||||
<span class="lineno">57</span> <span class="n">dropout_p</span><span class="o">=</span><span class="mf">0.0</span><span class="p">,</span>
|
||||
<span class="lineno">58</span> <span class="n">is_causal</span><span class="o">=</span><span class="kc">True</span><span class="p">,</span> <span class="c1"># for the triangular mask</span>
|
||||
<span class="lineno">59</span> <span class="p">)</span>
|
||||
<span class="lineno">60</span>
|
||||
<span class="lineno">61</span> <span class="n">attn_output</span> <span class="o">=</span> <span class="n">attn_output</span><span class="o">.</span><span class="n">transpose</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="mi">2</span><span class="p">)</span><span class="o">.</span><span class="n">contiguous</span><span class="p">()</span>
|
||||
<span class="lineno">62</span> <span class="n">attn_output</span> <span class="o">=</span> <span class="n">attn_output</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="n">batch_size</span><span class="p">,</span> <span class="n">seq_length</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">embed_dim</span><span class="p">)</span>
|
||||
<span class="lineno">63</span>
|
||||
<span class="lineno">64</span> <span class="n">attn_output</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">c_proj</span><span class="p">(</span><span class="n">attn_output</span><span class="p">)</span>
|
||||
<span class="lineno">65</span>
|
||||
<span class="lineno">66</span> <span class="k">return</span> <span class="n">attn_output</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">68</span> <span class="k">def</span> <span class="nf">_split_heads</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-13'>
|
||||
@ -256,10 +249,12 @@
|
||||
<div class='section-link'>
|
||||
<a href='#section-13'>#</a>
|
||||
</div>
|
||||
|
||||
<p>Split last dimension to <code class="highlight"><span></span><span class="p">[</span><span class="n">n_heads</span><span class="p">,</span> <span class="n">d_head</span><span class="p">]</span></code>
|
||||
</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">69</span><span class="k">class</span> <span class="nc">Block</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 class="highlight"><pre><span class="lineno">73</span> <span class="n">x</span> <span class="o">=</span> <span class="n">x</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="n">x</span><span class="o">.</span><span class="n">shape</span><span class="p">[:</span><span class="o">-</span><span class="mi">1</span><span class="p">]</span> <span class="o">+</span> <span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">n_heads</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">d_head</span><span class="p">))</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-14'>
|
||||
@ -267,38 +262,26 @@
|
||||
<div class='section-link'>
|
||||
<a href='#section-14'>#</a>
|
||||
</div>
|
||||
|
||||
<p>Reorder to <code class="highlight"><span></span><span class="p">[</span><span class="n">batch_size</span><span class="p">,</span> <span class="n">head</span><span class="p">,</span> <span class="n">seq_length</span><span class="p">,</span> <span class="n">d_head</span><span class="p">]</span></code>
|
||||
</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">70</span> <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">n_embed</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">layer_norm_epsilon</span><span class="p">:</span> <span class="nb">float</span><span class="p">,</span> <span class="n">r</span><span class="p">:</span> <span class="nb">int</span><span class="p">):</span>
|
||||
<span class="lineno">71</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
|
||||
<span class="lineno">72</span> <span class="bp">self</span><span class="o">.</span><span class="n">pre_norm</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">LayerNorm</span><span class="p">(</span><span class="n">n_embed</span><span class="p">,</span> <span class="n">eps</span><span class="o">=</span><span class="n">layer_norm_epsilon</span><span class="p">)</span>
|
||||
<span class="lineno">73</span> <span class="bp">self</span><span class="o">.</span><span class="n">attn</span> <span class="o">=</span> <span class="n">MultiHeadAttention</span><span class="p">(</span><span class="n">n_embed</span><span class="p">,</span> <span class="n">r</span><span class="p">)</span>
|
||||
<span class="lineno">74</span> <span class="bp">self</span><span class="o">.</span><span class="n">post_norm</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">LayerNorm</span><span class="p">(</span><span class="n">n_embed</span><span class="p">,</span> <span class="n">eps</span><span class="o">=</span><span class="n">layer_norm_epsilon</span><span class="p">)</span>
|
||||
<span class="lineno">75</span> <span class="bp">self</span><span class="o">.</span><span class="n">ffn</span> <span class="o">=</span> <span class="n">FFN</span><span class="p">(</span><span class="n">n_embed</span> <span class="o">*</span> <span class="mi">4</span><span class="p">,</span> <span class="n">n_embed</span><span class="p">,</span> <span class="n">r</span><span class="p">)</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">75</span> <span class="k">return</span> <span class="n">x</span><span class="o">.</span><span class="n">permute</span><span class="p">(</span><span class="mi">0</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">3</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-15'>
|
||||
<div class='docs'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-15'>#</a>
|
||||
</div>
|
||||
|
||||
<ul><li><code class="highlight"><span></span><span class="n">x</span></code>
|
||||
is the embeddings tensor with shape <code class="highlight"><span></span><span class="p">[</span><span class="n">batch_size</span><span class="p">,</span> <span class="n">seq_len</span><span class="p">,</span> <span class="n">d_model</span><span class="p">]</span></code>
|
||||
</li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">77</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">hidden_states</span><span class="p">):</span>
|
||||
<span class="lineno">78</span> <span class="n">residual</span> <span class="o">=</span> <span class="n">hidden_states</span>
|
||||
<span class="lineno">79</span> <span class="n">hidden_states</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">pre_norm</span><span class="p">(</span><span class="n">hidden_states</span><span class="p">)</span>
|
||||
<span class="lineno">80</span>
|
||||
<span class="lineno">81</span> <span class="n">attn_output</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">attn</span><span class="p">(</span><span class="n">hidden_states</span><span class="p">)</span>
|
||||
<span class="lineno">82</span>
|
||||
<span class="lineno">83</span> <span class="n">hidden_states</span> <span class="o">=</span> <span class="n">attn_output</span> <span class="o">+</span> <span class="n">residual</span>
|
||||
<span class="lineno">84</span> <span class="n">residual</span> <span class="o">=</span> <span class="n">hidden_states</span>
|
||||
<span class="lineno">85</span> <span class="n">hidden_states</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">post_norm</span><span class="p">(</span><span class="n">hidden_states</span><span class="p">)</span>
|
||||
<span class="lineno">86</span> <span class="n">feed_forward_output</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">ffn</span><span class="p">(</span><span class="n">hidden_states</span><span class="p">)</span>
|
||||
<span class="lineno">87</span> <span class="n">hidden_states</span> <span class="o">=</span> <span class="n">feed_forward_output</span> <span class="o">+</span> <span class="n">residual</span>
|
||||
<span class="lineno">88</span>
|
||||
<span class="lineno">89</span> <span class="k">return</span> <span class="n">hidden_states</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">77</span> <span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">)</span> <span class="o">-></span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">:</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-16'>
|
||||
@ -309,7 +292,7 @@
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">92</span><span class="k">class</span> <span class="nc">GPTModel</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 class="highlight"><pre><span class="lineno">81</span> <span class="n">batch_size</span><span class="p">,</span> <span class="n">seq_length</span><span class="p">,</span> <span class="n">_</span> <span class="o">=</span> <span class="n">x</span><span class="o">.</span><span class="n">shape</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-17'>
|
||||
@ -317,36 +300,27 @@
|
||||
<div class='section-link'>
|
||||
<a href='#section-17'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">93</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">layer_norm_epsilon</span><span class="p">:</span> <span class="nb">float</span><span class="p">,</span> <span class="n">n_embd</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">n_layer</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">n_positions</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span>
|
||||
<span class="lineno">94</span> <span class="n">vocab_size</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">r</span><span class="p">:</span> <span class="nb">int</span><span class="p">):</span>
|
||||
<span class="lineno">95</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
|
||||
<span class="lineno">96</span>
|
||||
<span class="lineno">97</span> <span class="bp">self</span><span class="o">.</span><span class="n">token_embedding</span> <span class="o">=</span> <span class="n">Embedding</span><span class="p">(</span><span class="n">vocab_size</span><span class="p">,</span> <span class="n">n_embd</span><span class="p">,</span> <span class="n">r</span><span class="o">=</span><span class="n">r</span><span class="p">)</span>
|
||||
<span class="lineno">98</span> <span class="bp">self</span><span class="o">.</span><span class="n">position_embedding</span> <span class="o">=</span> <span class="n">Embedding</span><span class="p">(</span><span class="n">n_positions</span><span class="p">,</span> <span class="n">n_embd</span><span class="p">,</span> <span class="n">r</span><span class="o">=</span><span class="n">r</span><span class="p">)</span>
|
||||
<span class="lineno">99</span>
|
||||
<span class="lineno">100</span> <span class="bp">self</span><span class="o">.</span><span class="n">blocks</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">ModuleList</span><span class="p">([</span><span class="n">Block</span><span class="p">(</span><span class="n">n_embd</span><span class="p">,</span> <span class="n">layer_norm_epsilon</span><span class="p">,</span> <span class="n">r</span><span class="o">=</span><span class="n">r</span><span class="p">)</span>
|
||||
<span class="lineno">101</span> <span class="k">for</span> <span class="n">_</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="n">n_layer</span><span class="p">)])</span>
|
||||
<span class="lineno">102</span>
|
||||
<span class="lineno">103</span> <span class="bp">self</span><span class="o">.</span><span class="n">final_norm</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">LayerNorm</span><span class="p">(</span><span class="n">n_embd</span><span class="p">,</span> <span class="n">eps</span><span class="o">=</span><span class="n">layer_norm_epsilon</span><span class="p">)</span>
|
||||
<span class="lineno">104</span>
|
||||
<span class="lineno">105</span> <span class="bp">self</span><span class="o">.</span><span class="n">lm_head</span> <span class="o">=</span> <span class="n">Linear</span><span class="p">(</span><span class="n">n_embd</span><span class="p">,</span> <span class="n">vocab_size</span><span class="p">,</span> <span class="n">r</span><span class="o">=</span><span class="n">r</span><span class="p">,</span> <span class="n">bias</span><span class="o">=</span><span class="kc">False</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-18'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-18'>#</a>
|
||||
</div>
|
||||
<ul><li><code class="highlight"><span></span><span class="n">input_ids</span></code>
|
||||
has shape <code class="highlight"><span></span><span class="p">[</span><span class="n">batch_size</span><span class="p">,</span> <span class="n">seq_len</span><span class="p">]</span></code>
|
||||
</li></ul>
|
||||
<p>Get query, key and value </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">107</span> <span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">input_ids</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 class="highlight"><pre><span class="lineno">84</span> <span class="n">q</span><span class="p">,</span> <span class="n">k</span><span class="p">,</span> <span class="n">v</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">qkv_projection</span><span class="p">(</span><span class="n">x</span><span class="p">)</span><span class="o">.</span><span class="n">split</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">d_model</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-18'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-18'>#</a>
|
||||
</div>
|
||||
<p>Transform them from shape <code class="highlight"><span></span><span class="p">[</span><span class="n">batch_size</span><span class="p">,</span> <span class="n">seq_len</span><span class="p">,</span> <span class="n">d_model</span><span class="p">]</span></code>
|
||||
to <code class="highlight"><span></span><span class="p">[</span><span class="n">batch_size</span><span class="p">,</span> <span class="n">head</span><span class="p">,</span> <span class="n">seq_length</span><span class="p">,</span> <span class="n">d_head</span><span class="p">]</span></code>
|
||||
</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">87</span> <span class="n">q</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">_split_heads</span><span class="p">(</span><span class="n">q</span><span class="p">)</span>
|
||||
<span class="lineno">88</span> <span class="n">k</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">_split_heads</span><span class="p">(</span><span class="n">k</span><span class="p">)</span>
|
||||
<span class="lineno">89</span> <span class="n">v</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">_split_heads</span><span class="p">(</span><span class="n">v</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-19'>
|
||||
@ -354,10 +328,11 @@
|
||||
<div class='section-link'>
|
||||
<a href='#section-19'>#</a>
|
||||
</div>
|
||||
|
||||
<p>Apply causal attention </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">111</span> <span class="n">batch_size</span><span class="p">,</span> <span class="n">seq_len</span> <span class="o">=</span> <span class="n">input_ids</span><span class="o">.</span><span class="n">shape</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">92</span> <span class="n">attn_output</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">nn</span><span class="o">.</span><span class="n">functional</span><span class="o">.</span><span class="n">scaled_dot_product_attention</span><span class="p">(</span><span class="n">q</span><span class="p">,</span> <span class="n">k</span><span class="p">,</span> <span class="n">v</span><span class="p">,</span> <span class="n">is_causal</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-20'>
|
||||
@ -365,11 +340,13 @@
|
||||
<div class='section-link'>
|
||||
<a href='#section-20'>#</a>
|
||||
</div>
|
||||
<p>Get token embeddings </p>
|
||||
<p>Transform them from shape <code class="highlight"><span></span><span class="p">[</span><span class="n">batch_size</span><span class="p">,</span> <span class="n">head</span><span class="p">,</span> <span class="n">seq_length</span><span class="p">,</span> <span class="n">d_head</span><span class="p">]</span></code>
|
||||
to <code class="highlight"><span></span><span class="p">[</span><span class="n">batch_size</span><span class="p">,</span> <span class="n">seq_len</span><span class="p">,</span> <span class="n">d_model</span><span class="p">]</span></code>
|
||||
</p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">114</span> <span class="n">token_embeddings</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">token_embedding</span><span class="p">(</span><span class="n">input_ids</span><span class="p">)</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">95</span> <span class="n">attn_output</span> <span class="o">=</span> <span class="n">attn_output</span><span class="o">.</span><span class="n">permute</span><span class="p">(</span><span class="mi">0</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">3</span><span class="p">)</span><span class="o">.</span><span class="n">reshape</span><span class="p">(</span><span class="n">batch_size</span><span class="p">,</span> <span class="n">seq_length</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">d_model</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-21'>
|
||||
@ -377,35 +354,42 @@
|
||||
<div class='section-link'>
|
||||
<a href='#section-21'>#</a>
|
||||
</div>
|
||||
<p>Get position ids </p>
|
||||
<p>Final project </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">116</span> <span class="n">position_ids</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">arange</span><span class="p">(</span><span class="n">seq_len</span><span class="p">,</span> <span class="n">device</span><span class="o">=</span><span class="n">input_ids</span><span class="o">.</span><span class="n">device</span><span class="p">)[</span><span class="kc">None</span><span class="p">,</span> <span class="p">:]</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">98</span> <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">output_projection</span><span class="p">(</span><span class="n">attn_output</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-22'>
|
||||
<div class='docs'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-22'>#</a>
|
||||
</div>
|
||||
<p>Get position embeddings </p>
|
||||
<h3>Decoder block</h3>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">118</span> <span class="n">position_embeddings</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">position_embedding</span><span class="p">(</span><span class="n">position_ids</span><span class="p">)</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">101</span><span class="k">class</span> <span class="nc">Block</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-23'>
|
||||
<div class='docs'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-23'>#</a>
|
||||
</div>
|
||||
<p>Add position embeddings </p>
|
||||
<ul><li><code class="highlight"><span></span><span class="n">d_model</span></code>
|
||||
is the number of dimensions in the embeddings </li>
|
||||
<li><code class="highlight"><span></span><span class="n">n_heads</span></code>
|
||||
is the number of heads </li>
|
||||
<li><code class="highlight"><span></span><span class="n">layer_norm_epsilon</span></code>
|
||||
is the layer norm epsilon </li>
|
||||
<li><code class="highlight"><span></span><span class="n">r</span></code>
|
||||
is the lora rank</li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">121</span> <span class="n">x</span> <span class="o">=</span> <span class="n">token_embeddings</span> <span class="o">+</span> <span class="n">position_embeddings</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">106</span> <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">d_model</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">n_heads</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">layer_norm_epsilon</span><span class="p">:</span> <span class="nb">float</span><span class="p">,</span> <span class="n">r</span><span class="p">:</span> <span class="nb">int</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-24'>
|
||||
@ -413,12 +397,10 @@
|
||||
<div class='section-link'>
|
||||
<a href='#section-24'>#</a>
|
||||
</div>
|
||||
<p>Run through transformer blocks </p>
|
||||
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">124</span> <span class="k">for</span> <span class="n">block</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">blocks</span><span class="p">:</span>
|
||||
<span class="lineno">125</span> <span class="n">x</span> <span class="o">=</span> <span class="n">block</span><span class="p">(</span><span class="n">x</span><span class="p">)</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">113</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-25'>
|
||||
@ -426,11 +408,11 @@
|
||||
<div class='section-link'>
|
||||
<a href='#section-25'>#</a>
|
||||
</div>
|
||||
<p>Final normalization </p>
|
||||
<p>Attention pre-normalization layer </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">128</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">final_norm</span><span class="p">(</span><span class="n">x</span><span class="p">)</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">115</span> <span class="bp">self</span><span class="o">.</span><span class="n">attn_norm</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">LayerNorm</span><span class="p">(</span><span class="n">d_model</span><span class="p">,</span> <span class="n">eps</span><span class="o">=</span><span class="n">layer_norm_epsilon</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-26'>
|
||||
@ -438,11 +420,287 @@
|
||||
<div class='section-link'>
|
||||
<a href='#section-26'>#</a>
|
||||
</div>
|
||||
<p>Attention layer </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">117</span> <span class="bp">self</span><span class="o">.</span><span class="n">attn</span> <span class="o">=</span> <span class="n">MultiHeadAttention</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">r</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>FFN pre-normalization layer </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">119</span> <span class="bp">self</span><span class="o">.</span><span class="n">ffn_norm</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">LayerNorm</span><span class="p">(</span><span class="n">d_model</span><span class="p">,</span> <span class="n">eps</span><span class="o">=</span><span class="n">layer_norm_epsilon</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>Feed-forward network </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">121</span> <span class="bp">self</span><span class="o">.</span><span class="n">ffn</span> <span class="o">=</span> <span class="n">FFN</span><span class="p">(</span><span class="n">d_model</span><span class="p">,</span> <span class="n">d_model</span> <span class="o">*</span> <span class="mi">4</span><span class="p">,</span> <span class="n">r</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-29'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-29'>#</a>
|
||||
</div>
|
||||
<ul><li><code class="highlight"><span></span><span class="n">x</span></code>
|
||||
is the embeddings tensor with shape <code class="highlight"><span></span><span class="p">[</span><span class="n">batch_size</span><span class="p">,</span> <span class="n">seq_len</span><span class="p">,</span> <span class="n">d_model</span><span class="p">]</span></code>
|
||||
</li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">123</span> <span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">)</span> <span class="o">-></span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">:</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-30'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-30'>#</a>
|
||||
</div>
|
||||
<p>Attention </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">128</span> <span class="n">x</span> <span class="o">=</span> <span class="n">x</span> <span class="o">+</span> <span class="bp">self</span><span class="o">.</span><span class="n">attn</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">attn_norm</span><span class="p">(</span><span class="n">x</span><span class="p">))</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-31'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-31'>#</a>
|
||||
</div>
|
||||
<p>FFN </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">130</span> <span class="n">x</span> <span class="o">=</span> <span class="n">x</span> <span class="o">+</span> <span class="bp">self</span><span class="o">.</span><span class="n">ffn</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">ffn_norm</span><span class="p">(</span><span class="n">x</span><span class="p">))</span>
|
||||
<span class="lineno">131</span>
|
||||
<span class="lineno">132</span> <span class="k">return</span> <span class="n">x</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-32'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-32'>#</a>
|
||||
</div>
|
||||
<h2>GPT2 Model</h2>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">135</span><span class="k">class</span> <span class="nc">GPTModel</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-33'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-33'>#</a>
|
||||
</div>
|
||||
<ul><li><code class="highlight"><span></span><span class="n">d_model</span></code>
|
||||
is the number of dimensions in the embeddings </li>
|
||||
<li><code class="highlight"><span></span><span class="n">n_heads</span></code>
|
||||
is the number of attention heads </li>
|
||||
<li><code class="highlight"><span></span><span class="n">n_layers</span></code>
|
||||
is the number of decoder layers </li>
|
||||
<li><code class="highlight"><span></span><span class="n">n_positions</span></code>
|
||||
is the number of positional embeddings </li>
|
||||
<li><code class="highlight"><span></span><span class="n">layer_norm_epsilon</span></code>
|
||||
is the layer norm epsilon </li>
|
||||
<li><code class="highlight"><span></span><span class="n">vocab_size</span></code>
|
||||
is the vocabulary size </li>
|
||||
<li><code class="highlight"><span></span><span class="n">r</span></code>
|
||||
is the lora rank</li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">140</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="o">*</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="lineno">141</span> <span class="n">n_heads</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">n_layers</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span>
|
||||
<span class="lineno">142</span> <span class="n">n_positions</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span>
|
||||
<span class="lineno">143</span> <span class="n">layer_norm_epsilon</span><span class="p">:</span> <span class="nb">float</span><span class="p">,</span>
|
||||
<span class="lineno">144</span> <span class="n">vocab_size</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">r</span><span class="p">:</span> <span class="nb">int</span><span class="p">):</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-34'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-34'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">154</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-35'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-35'>#</a>
|
||||
</div>
|
||||
<p>Token and absolute positional embeddings </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">157</span> <span class="bp">self</span><span class="o">.</span><span class="n">token_embedding</span> <span class="o">=</span> <span class="n">Embedding</span><span class="p">(</span><span class="n">vocab_size</span><span class="p">,</span> <span class="n">d_model</span><span class="p">,</span> <span class="n">r</span><span class="o">=</span><span class="n">r</span><span class="p">)</span>
|
||||
<span class="lineno">158</span> <span class="bp">self</span><span class="o">.</span><span class="n">position_embedding</span> <span class="o">=</span> <span class="n">Embedding</span><span class="p">(</span><span class="n">n_positions</span><span class="p">,</span> <span class="n">d_model</span><span class="p">,</span> <span class="n">r</span><span class="o">=</span><span class="n">r</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>Decoder blocks </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">161</span> <span class="bp">self</span><span class="o">.</span><span class="n">blocks</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">ModuleList</span><span class="p">([</span><span class="n">Block</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">layer_norm_epsilon</span><span class="p">,</span> <span class="n">r</span><span class="o">=</span><span class="n">r</span><span class="p">)</span>
|
||||
<span class="lineno">162</span> <span class="k">for</span> <span class="n">_</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="n">n_layers</span><span class="p">)])</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-37'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-37'>#</a>
|
||||
</div>
|
||||
<p>Final layer norm </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">165</span> <span class="bp">self</span><span class="o">.</span><span class="n">final_norm</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">LayerNorm</span><span class="p">(</span><span class="n">d_model</span><span class="p">,</span> <span class="n">eps</span><span class="o">=</span><span class="n">layer_norm_epsilon</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>Projection layer to logit space </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">167</span> <span class="bp">self</span><span class="o">.</span><span class="n">lm_head</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">vocab_size</span><span class="p">,</span> <span class="n">r</span><span class="o">=</span><span class="n">r</span><span class="p">,</span> <span class="n">bias</span><span class="o">=</span><span class="kc">False</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-39'>
|
||||
<div class='docs doc-strings'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-39'>#</a>
|
||||
</div>
|
||||
<ul><li><code class="highlight"><span></span><span class="n">input_ids</span></code>
|
||||
has shape <code class="highlight"><span></span><span class="p">[</span><span class="n">batch_size</span><span class="p">,</span> <span class="n">seq_len</span><span class="p">]</span></code>
|
||||
</li></ul>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">169</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">input_ids</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-40'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-40'>#</a>
|
||||
</div>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">173</span> <span class="n">batch_size</span><span class="p">,</span> <span class="n">seq_len</span> <span class="o">=</span> <span class="n">input_ids</span><span class="o">.</span><span class="n">shape</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>Get token embeddings </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">176</span> <span class="n">token_embeddings</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">token_embedding</span><span class="p">(</span><span class="n">input_ids</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-42'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-42'>#</a>
|
||||
</div>
|
||||
<p>Get position ids </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">178</span> <span class="n">position_ids</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">arange</span><span class="p">(</span><span class="n">seq_len</span><span class="p">,</span> <span class="n">device</span><span class="o">=</span><span class="n">input_ids</span><span class="o">.</span><span class="n">device</span><span class="p">)[</span><span class="kc">None</span><span class="p">,</span> <span class="p">:]</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-43'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-43'>#</a>
|
||||
</div>
|
||||
<p>Get position embeddings </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">180</span> <span class="n">position_embeddings</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">position_embedding</span><span class="p">(</span><span class="n">position_ids</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>Add position embeddings </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">183</span> <span class="n">x</span> <span class="o">=</span> <span class="n">token_embeddings</span> <span class="o">+</span> <span class="n">position_embeddings</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>Run through transformer blocks </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">186</span> <span class="k">for</span> <span class="n">block</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">blocks</span><span class="p">:</span>
|
||||
<span class="lineno">187</span> <span class="n">x</span> <span class="o">=</span> <span class="n">block</span><span class="p">(</span><span class="n">x</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='section' id='section-46'>
|
||||
<div class='docs'>
|
||||
<div class='section-link'>
|
||||
<a href='#section-46'>#</a>
|
||||
</div>
|
||||
<p>Final normalization </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">190</span> <span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">final_norm</span><span class="p">(</span><span class="n">x</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>Get logits from projection layer </p>
|
||||
|
||||
</div>
|
||||
<div class='code'>
|
||||
<div class="highlight"><pre><span class="lineno">130</span> <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">lm_head</span><span class="p">(</span><span class="n">x</span><span class="p">)</span></pre></div>
|
||||
<div class="highlight"><pre><span class="lineno">192</span> <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">lm_head</span><span class="p">(</span><span class="n">x</span><span class="p">)</span></pre></div>
|
||||
</div>
|
||||
</div>
|
||||
<div class='footer'>
|
||||
|
||||
File diff suppressed because one or more lines are too long
@ -1296,21 +1296,21 @@
|
||||
|
||||
<url>
|
||||
<loc>https://nn.labml.ai/lora/gpt2.html</loc>
|
||||
<lastmod>2024-08-16T16:30:00+00:00</lastmod>
|
||||
<lastmod>2024-08-18T16:30:00+00:00</lastmod>
|
||||
<priority>1.00</priority>
|
||||
</url>
|
||||
|
||||
|
||||
<url>
|
||||
<loc>https://nn.labml.ai/lora/index.html</loc>
|
||||
<lastmod>2024-08-03T16:30:00+00:00</lastmod>
|
||||
<lastmod>2024-08-23T16:30:00+00:00</lastmod>
|
||||
<priority>1.00</priority>
|
||||
</url>
|
||||
|
||||
|
||||
<url>
|
||||
<loc>https://nn.labml.ai/lora/experiment.html</loc>
|
||||
<lastmod>2024-08-16T16:30:00+00:00</lastmod>
|
||||
<lastmod>2024-08-23T16:30:00+00:00</lastmod>
|
||||
<priority>1.00</priority>
|
||||
</url>
|
||||
|
||||
|
||||
Reference in New Issue
Block a user