Buckets:
| <meta charset="utf-8" /><meta name="hf:doc:metadata" content="{"title":"Customizing model components","local":"customizing-model-components","sections":[{"title":"Attention class","local":"attention-class","sections":[],"depth":2},{"title":"LoRA","local":"lora","sections":[],"depth":2}],"depth":1}"> | |
| <link href="/docs/transformers/pr_33892/en/_app/immutable/assets/0.e3b0c442.css" rel="modulepreload"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33892/en/_app/immutable/entry/start.b2c4257a.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33892/en/_app/immutable/chunks/scheduler.31fdf58d.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33892/en/_app/immutable/chunks/singletons.9860629f.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33892/en/_app/immutable/chunks/index.252883d5.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33892/en/_app/immutable/chunks/paths.e85c0ec8.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33892/en/_app/immutable/entry/app.05ef1f97.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33892/en/_app/immutable/chunks/preload-helper.40847a0e.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33892/en/_app/immutable/chunks/index.2f76fdf0.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33892/en/_app/immutable/nodes/0.ca4aafa4.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33892/en/_app/immutable/chunks/each.e59479a4.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33892/en/_app/immutable/nodes/30.c910d222.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33892/en/_app/immutable/chunks/CopyLLMTxtMenu.ff482081.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33892/en/_app/immutable/chunks/MermaidChart.svelte_svelte_type_style_lang.71f274cc.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33892/en/_app/immutable/chunks/IconCopy.ac192424.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33892/en/_app/immutable/chunks/CodeBlock.ab12f8e1.js"><!-- HEAD_svelte-u9bgzb_START --><meta name="hf:doc:metadata" content="{"title":"Customizing model components","local":"customizing-model-components","sections":[{"title":"Attention class","local":"attention-class","sections":[],"depth":2},{"title":"LoRA","local":"lora","sections":[],"depth":2}],"depth":1}"><!-- HEAD_svelte-u9bgzb_END --> <p></p> <div class="items-center shrink-0 min-w-[100px] max-sm:min-w-[50px] justify-end ml-auto flex" style="float: right; margin-left: 10px; display: inline-flex; position: relative; z-index: 10;"><div class="inline-flex rounded-md max-sm:rounded-sm"><button class="inline-flex items-center gap-1 max-sm:gap-0.5 h-6 max-sm:h-5 px-2 max-sm:px-1.5 text-[11px] max-sm:text-[9px] font-medium text-gray-800 border border-r-0 rounded-l-md max-sm:rounded-l-sm border-gray-200 bg-white hover:shadow-inner dark:border-gray-850 dark:bg-gray-950 dark:text-gray-200 dark:hover:bg-gray-800" aria-live="polite"><span class="inline-flex items-center justify-center rounded-md p-0.5 max-sm:p-0"><svg class="w-3 h-3 max-sm:w-2.5 max-sm:h-2.5" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg></span> <span>Copy page</span></button> <button class="inline-flex items-center justify-center w-6 max-sm:w-5 h-6 max-sm:h-5 disabled:pointer-events-none text-sm text-gray-500 hover:text-gray-700 dark:hover:text-white rounded-r-md max-sm:rounded-r-sm border border-l transition border-gray-200 bg-white hover:shadow-inner dark:border-gray-850 dark:bg-gray-950 dark:text-gray-200 dark:hover:bg-gray-800" aria-haspopup="menu" aria-expanded="false" aria-label="Open copy menu"><svg class="transition-transform text-gray-400 overflow-visible w-3 h-3 max-sm:w-2.5 max-sm:h-2.5 rotate-0" width="1em" height="1em" viewBox="0 0 12 7" fill="none" xmlns="http://www.w3.org/2000/svg"><path d="M1 1L6 6L11 1" stroke="currentColor"></path></svg></button></div> </div> <h1 class="relative group"><a id="customizing-model-components" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#customizing-model-components"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>Customizing model components</span></h1> <p data-svelte-h="svelte-1iy0nzn">Another way to customize a model is to modify their components, rather than writing a new model entirely, allowing you to tailor a model to your specific use case. For example, you can add new layers or optimize the attention mechanism of an architecture. Customizations are applied directly to a Transformers model so that you can continue to use features such as <a href="/docs/transformers/pr_33892/en/main_classes/trainer#transformers.Trainer">Trainer</a>, <a href="/docs/transformers/pr_33892/en/main_classes/model#transformers.PreTrainedModel">PreTrainedModel</a>, and the <a href="https://huggingface.co/docs/peft/en/index" rel="nofollow">PEFT</a> library.</p> <p data-svelte-h="svelte-1rh1a31">This guide will show you how to customize a models attention mechanism in order to apply <a href="https://huggingface.co/docs/peft/conceptual_guides/adapter#low-rank-adaptation-lora" rel="nofollow">Low-Rank Adaptation (LoRA)</a> to it.</p> <blockquote class="tip"><p data-svelte-h="svelte-13tb4k6">The <a href="https://github.com/huggingface/transformers/blob/9985d06add07a4cc691dc54a7e34f54205c04d40/src/transformers/utils/import_utils.py#L2286" rel="nofollow">clear_import_cache</a> utility is very useful when you’re iteratively modifying and developing model code. It removes all cached Transformers modules and allows Python to reload the modified code without constantly restarting your environment.</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoModel | |
| <span class="hljs-keyword">from</span> transformers.utils.import_utils <span class="hljs-keyword">import</span> clear_import_cache | |
| model = AutoModel.from_pretrained(<span class="hljs-string">"bert-base-uncased"</span>) | |
| <span class="hljs-comment"># modifications to model code</span> | |
| <span class="hljs-comment"># clear cache to reload modified code</span> | |
| clear_import_cache() | |
| <span class="hljs-comment"># re-import to use updated code</span> | |
| model = AutoModel.from_pretrained(<span class="hljs-string">"bert-base-uncased"</span>)<!-- HTML_TAG_END --></pre></div></blockquote> <h2 class="relative group"><a id="attention-class" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#attention-class"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>Attention class</span></h2> <p data-svelte-h="svelte-64c6cg"><a href="./model_doc/sam">Segment Anything</a> is an image segmentation model, and it combines the query-key-value (<code>qkv</code>) projection in its attention mechanisms. To reduce the number of trainable parameters and computational overhead, you can apply LoRA to the <code>qkv</code> projection. This requires splitting the <code>qkv</code> projection so that you can separately target the <code>q</code> and <code>v</code> with LoRA.</p> <ol data-svelte-h="svelte-1i23me3"><li>Create a custom attention class, <code>SamVisionAttentionSplit</code>, by subclassing the original <code>SamVisionAttention</code> class. In the <code>__init__</code>, delete the combined <code>qkv</code> and create a separate linear layer for <code>q</code>, <code>k</code> and <code>v</code>.</li></ol> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-keyword">import</span> torch | |
| <span class="hljs-keyword">import</span> torch.nn <span class="hljs-keyword">as</span> nn | |
| <span class="hljs-keyword">from</span> transformers.models.sam.modeling_sam <span class="hljs-keyword">import</span> SamVisionAttention | |
| <span class="hljs-keyword">class</span> <span class="hljs-title class_">SamVisionAttentionSplit</span>(SamVisionAttention, nn.Module): | |
| <span class="hljs-keyword">def</span> <span class="hljs-title function_">__init__</span>(<span class="hljs-params">self, config, window_size</span>): | |
| <span class="hljs-built_in">super</span>().__init__(config, window_size) | |
| <span class="hljs-comment"># remove combined qkv</span> | |
| <span class="hljs-keyword">del</span> self.qkv | |
| <span class="hljs-comment"># separate q, k, v projections</span> | |
| self.q = nn.Linear(config.hidden_size, config.hidden_size, bias=config.qkv_bias) | |
| self.k = nn.Linear(config.hidden_size, config.hidden_size, bias=config.qkv_bias) | |
| self.v = nn.Linear(config.hidden_size, config.hidden_size, bias=config.qkv_bias) | |
| self._register_load_state_dict_pre_hook(self.split_q_k_v_load_hook)<!-- HTML_TAG_END --></pre></div> <ol start="2" data-svelte-h="svelte-11rant"><li>The <code>_split_qkv_load_hook</code> function splits the pretrained <code>qkv</code> weights into separate <code>q</code>, <code>k</code>, and <code>v</code> weights when loading the model to ensure compatibility with any pretrained model.</li></ol> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --> <span class="hljs-keyword">def</span> <span class="hljs-title function_">split_q_k_v_load_hook</span>(<span class="hljs-params">self, state_dict, prefix, *args</span>): | |
| keys_to_delete = [] | |
| <span class="hljs-keyword">for</span> key <span class="hljs-keyword">in</span> <span class="hljs-built_in">list</span>(state_dict.keys()): | |
| <span class="hljs-keyword">if</span> <span class="hljs-string">"qkv."</span> <span class="hljs-keyword">in</span> key: | |
| <span class="hljs-comment"># split q, k, v from the combined projection</span> | |
| q, k, v = state_dict[key].chunk(<span class="hljs-number">3</span>, dim=<span class="hljs-number">0</span>) | |
| <span class="hljs-comment"># replace with individual q, k, v projections</span> | |
| state_dict[key.replace(<span class="hljs-string">"qkv."</span>, <span class="hljs-string">"q."</span>)] = q | |
| state_dict[key.replace(<span class="hljs-string">"qkv."</span>, <span class="hljs-string">"k."</span>)] = k | |
| state_dict[key.replace(<span class="hljs-string">"qkv."</span>, <span class="hljs-string">"v."</span>)] = v | |
| <span class="hljs-comment"># mark the old qkv key for deletion</span> | |
| keys_to_delete.append(key) | |
| <span class="hljs-comment"># remove old qkv keys</span> | |
| <span class="hljs-keyword">for</span> key <span class="hljs-keyword">in</span> keys_to_delete: | |
| <span class="hljs-keyword">del</span> state_dict[key]<!-- HTML_TAG_END --></pre></div> <ol start="3" data-svelte-h="svelte-6hmxrg"><li>In the <code>forward</code> pass, <code>q</code>, <code>k</code>, and <code>v</code> are computed separately while the rest of the attention mechanism remains the same.</li></ol> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --> <span class="hljs-keyword">def</span> <span class="hljs-title function_">forward</span>(<span class="hljs-params">self, hidden_states: torch.Tensor, output_attentions=<span class="hljs-literal">False</span></span>) -> torch.Tensor: | |
| batch_size, height, width, _ = hidden_states.shape | |
| qkv_shapes = (batch_size * self.num_attention_heads, height * width, -<span class="hljs-number">1</span>) | |
| query = self.q(hidden_states).reshape((batch_size, height * width,self.num_attention_heads, -<span class="hljs-number">1</span>)).permute(<span class="hljs-number">0</span>,<span class="hljs-number">2</span>,<span class="hljs-number">1</span>,<span class="hljs-number">3</span>).reshape(qkv_shapes) | |
| key = self.k(hidden_states).reshape((batch_size, height * width,self.num_attention_heads, -<span class="hljs-number">1</span>)).permute(<span class="hljs-number">0</span>,<span class="hljs-number">2</span>,<span class="hljs-number">1</span>,<span class="hljs-number">3</span>).reshape(qkv_shapes) | |
| value = self.v(hidden_states).reshape((batch_size, height * width,self.num_attention_heads, -<span class="hljs-number">1</span>)).permute(<span class="hljs-number">0</span>,<span class="hljs-number">2</span>,<span class="hljs-number">1</span>,<span class="hljs-number">3</span>).reshape(qkv_shapes) | |
| attn_weights = (query * self.scale) @ key.transpose(-<span class="hljs-number">2</span>, -<span class="hljs-number">1</span>) | |
| attn_weights = torch.nn.functional.softmax(attn_weights, dtype=torch.float32, dim=-<span class="hljs-number">1</span>).to(query.dtype) | |
| attn_probs = nn.functional.dropout(attn_weights, p=self.dropout, training=self.training) | |
| attn_output = (attn_probs @ value).reshape(batch_size, self.num_attention_heads, height, width, -<span class="hljs-number">1</span>) | |
| attn_output = attn_output.permute(<span class="hljs-number">0</span>, <span class="hljs-number">2</span>, <span class="hljs-number">3</span>, <span class="hljs-number">1</span>, <span class="hljs-number">4</span>).reshape(batch_size, height, width, -<span class="hljs-number">1</span>) | |
| attn_output = self.proj(attn_output) | |
| <span class="hljs-keyword">if</span> output_attentions: | |
| outputs = (attn_output, attn_weights) | |
| <span class="hljs-keyword">else</span>: | |
| outputs = (attn_output, <span class="hljs-literal">None</span>) | |
| <span class="hljs-keyword">return</span> outputs<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-nkh0b4">Assign the custom <code>SamVisionAttentionSplit</code> class to the original models <code>SamVisionAttention</code> module to replace it. All instances of <code>SamVisionAttention</code> in the model is replaced with the split attention version.</p> <p data-svelte-h="svelte-fcgjrn">Load the model with <a href="/docs/transformers/pr_33892/en/main_classes/model#transformers.PreTrainedModel.from_pretrained">from_pretrained()</a>.</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> SamModel | |
| <span class="hljs-comment"># load the pretrained SAM model</span> | |
| model = SamModel.from_pretrained(<span class="hljs-string">"facebook/sam-vit-base"</span>) | |
| <span class="hljs-comment"># replace the attention class in the vision_encoder module</span> | |
| <span class="hljs-keyword">for</span> layer <span class="hljs-keyword">in</span> model.vision_encoder.layers: | |
| <span class="hljs-keyword">if</span> <span class="hljs-built_in">hasattr</span>(layer, <span class="hljs-string">"attn"</span>): | |
| layer.attn = SamVisionAttentionSplit(model.config.vision_config, model.config.vision_config.window_size)<!-- HTML_TAG_END --></pre></div> <h2 class="relative group"><a id="lora" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#lora"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>LoRA</span></h2> <p data-svelte-h="svelte-1km26rk">With separate <code>q</code>, <code>k</code>, and <code>v</code> projections, apply LoRA to <code>q</code> and <code>v</code>.</p> <p data-svelte-h="svelte-1hvvkbj">Create a <a href="https://huggingface.co/docs/peft/package_reference/config#peft.PeftConfig" rel="nofollow">LoraConfig</a> and specify the rank <code>r</code>, <code>lora_alpha</code>, <code>lora_dropout</code>, <code>task_type</code>, and most importantly, the modules to target.</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> peft <span class="hljs-keyword">import</span> LoraConfig, get_peft_model | |
| config = LoraConfig( | |
| r=<span class="hljs-number">16</span>, | |
| lora_alpha=<span class="hljs-number">32</span>, | |
| <span class="hljs-comment"># apply LoRA to q and v</span> | |
| target_modules=[<span class="hljs-string">"q"</span>, <span class="hljs-string">"v"</span>], | |
| lora_dropout=<span class="hljs-number">0.1</span>, | |
| task_type=<span class="hljs-string">"FEATURE_EXTRACTION"</span> | |
| )<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1gzm1t1">Pass the model and <a href="https://huggingface.co/docs/peft/package_reference/config#peft.PeftConfig" rel="nofollow">LoraConfig</a> to <a href="https://huggingface.co/docs/peft/package_reference/peft_model#peft.get_peft_model" rel="nofollow">get_peft_model</a> to apply LoRA to the model.</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->model = get_peft_model(model, config)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1ctingj">Call <a href="https://huggingface.co/docs/peft/package_reference/peft_model#peft.PeftMixedModel.print_trainable_parameters" rel="nofollow">print_trainable_parameters</a> to view the number of parameters you’re training as a result versus the total number of parameters.</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->model.print_trainable_parameters() | |
| <span class="hljs-string">"trainable params: 589,824 || all params: 94,274,096 || trainable%: 0.6256"</span><!-- HTML_TAG_END --></pre></div> <a class="!text-gray-400 !no-underline text-sm flex items-center not-prose mt-4" href="https://github.com/huggingface/transformers/blob/main/docs/source/en/how_to_hack_models.md" target="_blank"><svg class="mr-1" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M31,16l-7,7l-1.41-1.41L28.17,16l-5.58-5.59L24,9l7,7z"></path><path d="M1,16l7-7l1.41,1.41L3.83,16l5.58,5.59L8,23l-7-7z"></path><path d="M12.419,25.484L17.639,6.552l1.932,0.518L14.351,26.002z"></path></svg> <span data-svelte-h="svelte-zjs2n5"><span class="underline">Update</span> on GitHub</span></a> <p></p> | |
| <script> | |
| { | |
| __sveltekit_16tnnm8 = { | |
| assets: "/docs/transformers/pr_33892/en", | |
| base: "/docs/transformers/pr_33892/en", | |
| env: {} | |
| }; | |
| const element = document.currentScript.parentElement; | |
| const data = [null,null]; | |
| Promise.all([ | |
| import("/docs/transformers/pr_33892/en/_app/immutable/entry/start.b2c4257a.js"), | |
| import("/docs/transformers/pr_33892/en/_app/immutable/entry/app.05ef1f97.js") | |
| ]).then(([kit, app]) => { | |
| kit.start(app, element, { | |
| node_ids: [0, 30], | |
| data, | |
| form: null, | |
| error: null | |
| }); | |
| }); | |
| } | |
| </script> | |
Xet Storage Details
- Size:
- 30.9 kB
- Xet hash:
- 94602db684f8b74ec7adf7b9fefe380044dcebf4cfd953229c44c27cd1cc45e8
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.