Buckets:
| <meta charset="utf-8" /><meta name="hf:doc:metadata" content="{"title":"Tracing model intermediate outputs","local":"tracing-model-intermediate-outputs","sections":[{"title":"Declare which submodules to capture","local":"declare-which-submodules-to-capture","sections":[],"depth":2},{"title":"Map output fields to submodules","local":"map-output-fields-to-submodules","sections":[],"depth":2},{"title":"Fine-grained control with OutputRecorder","local":"fine-grained-control-with-outputrecorder","sections":[],"depth":2},{"title":"Patch layer classes","local":"patch-layer-classes","sections":[],"depth":2}],"depth":1}"> | |
| <link href="/docs/transformers/pr_36895/en/_app/immutable/assets/0.e3b0c442.css" rel="modulepreload"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_36895/en/_app/immutable/entry/start.b88698fb.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_36895/en/_app/immutable/chunks/scheduler.31fdf58d.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_36895/en/_app/immutable/chunks/singletons.18caffe4.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_36895/en/_app/immutable/chunks/index.252883d5.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_36895/en/_app/immutable/chunks/paths.a82d371a.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_36895/en/_app/immutable/entry/app.ac300a2c.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_36895/en/_app/immutable/chunks/preload-helper.643b8c40.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_36895/en/_app/immutable/chunks/index.2f76fdf0.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_36895/en/_app/immutable/nodes/0.1669b2ea.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_36895/en/_app/immutable/chunks/each.e59479a4.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_36895/en/_app/immutable/nodes/579.53250f8d.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_36895/en/_app/immutable/chunks/CopyLLMTxtMenu.9cbc081e.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_36895/en/_app/immutable/chunks/MermaidChart.svelte_svelte_type_style_lang.945f6700.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_36895/en/_app/immutable/chunks/IconCopy.ac192424.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_36895/en/_app/immutable/chunks/CodeBlock.e52df5d6.js"><!-- HEAD_svelte-u9bgzb_START --><meta name="hf:doc:metadata" content="{"title":"Tracing model intermediate outputs","local":"tracing-model-intermediate-outputs","sections":[{"title":"Declare which submodules to capture","local":"declare-which-submodules-to-capture","sections":[],"depth":2},{"title":"Map output fields to submodules","local":"map-output-fields-to-submodules","sections":[],"depth":2},{"title":"Fine-grained control with OutputRecorder","local":"fine-grained-control-with-outputrecorder","sections":[],"depth":2},{"title":"Patch layer classes","local":"patch-layer-classes","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 h-7 max-sm:h-7 px-2 max-sm:px-1.5 text-sm 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 hover:text-gray-800 dark:hover:text-gray-200"><svg class="sm:size-3.5 size-3" 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-7 max-sm:h-7 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 sm:size-3.5 size-3 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="tracing-model-intermediate-outputs" 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="#tracing-model-intermediate-outputs"><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>Tracing model intermediate outputs</span></h1> <p data-svelte-h="svelte-1vdtp15">Every model’s <code>forward()</code> method used to manually resolve <code>None</code> flags like <code>output_attentions</code> from config defaults, accumulate per-layer attention weights and hidden states into tuples, and convert <code>ModelOutput</code> dataclasses to plain tuples when <code>return_dict=False</code>. Two decorators replace all of that boilerplate.</p> <ul data-svelte-h="svelte-1pui02c"><li><code>@capture_outputs</code> resolves output flags, collects intermediate values, and handles <code>return_dict</code> conversion.</li> <li><code>@merge_with_config_defaults</code> resolves <code>use_cache</code> from config. Omit it for models that don’t cache, like <a href="/docs/transformers/pr_36895/en/model_doc/clip#transformers.CLIPModel">CLIPModel</a>.</li></ul> <p data-svelte-h="svelte-e5sa0c">You’ll mostly encounter these decorators when integrating a new model. See [adding a model to 🤗 Transformers]./modular_transformers) for a step-by-step guide.</p> <h2 class="relative group"><a id="declare-which-submodules-to-capture" 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="#declare-which-submodules-to-capture"><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>Declare which submodules to capture</span></h2> <p data-svelte-h="svelte-12wgehg">Apply <code>@capture_outputs</code> to the base model’s <code>forward()</code> method. It attaches forward hooks that:</p> <ul data-svelte-h="svelte-she9ay"><li>Intercept outputs from specified submodule classes during the forward pass, without those submodules needing to know they’re being observed.</li> <li>Collect per-layer attention weights and hidden states into tuples.</li> <li>Inject collected values into the returned <code>ModelOutput</code> dataclass.</li> <li>Convert the dataclass to a plain tuple when <code>return_dict=False</code>.</li> <li>Resolve <code>output_attentions</code> and <code>output_hidden_states</code> from kwargs or <code>self.config</code> when <code>None</code>.</li></ul> <h2 class="relative group"><a id="map-output-fields-to-submodules" 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="#map-output-fields-to-submodules"><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>Map output fields to submodules</span></h2> <p data-svelte-h="svelte-1hrslsi"><code>@capture_outputs</code> needs to know which submodule produces which output. Declare a <code>_can_record_outputs</code> class-level dictionary on your <code>PreTrainedModel</code> subclass. Each key is an output field name (<code>"hidden_states"</code>, <code>"attentions"</code>, <code>"cross_attentions"</code>), and each value is a module class or an <code>OutputRecorder</code> instance.</p> <h2 class="relative group"><a id="fine-grained-control-with-outputrecorder" 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="#fine-grained-control-with-outputrecorder"><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>Fine-grained control with OutputRecorder</span></h2> <p data-svelte-h="svelte-1k176dm"><code>OutputRecorder</code> accepts a <code>target_class</code> (an <code>nn.Module</code> subclass whose outputs to collect) and an optional <code>index</code> to select which element of the module’s output tuple to get. Pass <code>layer_name</code> to attach hooks only to modules with a specific attribute name. Use <code>layer_name</code> when two layers share the same class, for example self-attention vs. cross-attention.</p> <p data-svelte-h="svelte-8nt2gz">The example below shows both decorators in practice with different levels of output control. See <a href="https://github.com/huggingface/transformers/blob/main/src/transformers/models/llama/modeling_llama.py" rel="nofollow">LlamaModel</a> for real world reference.</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="language-python "><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> ...processing_utils <span class="hljs-keyword">import</span> Unpack | |
| <span class="hljs-keyword">from</span> ...utils <span class="hljs-keyword">import</span> TransformersKwargs | |
| <span class="hljs-keyword">from</span> ...utils.generic <span class="hljs-keyword">import</span> merge_with_config_defaults | |
| <span class="hljs-keyword">from</span> ...utils.output_capturing <span class="hljs-keyword">import</span> capture_outputs, OutputRecorder | |
| <span class="hljs-keyword">class</span> <span class="hljs-title class_">MyPreTrainedModel</span>(<span class="hljs-title class_ inherited__">PreTrainedModel</span>): | |
| _can_record_outputs = { | |
| <span class="hljs-comment"># Capture hidden_states: hook fires after each MyDecoderBlock forward,</span> | |
| <span class="hljs-comment"># grabbing its first output (index 0 by default).</span> | |
| <span class="hljs-string">"hidden_states"</span>: MyDecoderBlock, | |
| <span class="hljs-comment"># Capture self-attention weights: hook fires after each MyAttention</span> | |
| <span class="hljs-comment"># forward, grabbing its second output (index=1 by default).</span> | |
| <span class="hljs-string">"attentions"</span>: MyAttention, | |
| <span class="hljs-comment"># Capture cross-attention weights: same class, different submodule.</span> | |
| <span class="hljs-comment"># layer_name targets the attribute `self.crossattention` inside the block.</span> | |
| <span class="hljs-comment"># Captures second output as requested (index=1)</span> | |
| <span class="hljs-string">"cross_attentions"</span>: OutputRecorder( | |
| MyAttention, layer_name=<span class="hljs-string">"crossattention"</span>, index=<span class="hljs-number">1</span> | |
| ), | |
| } | |
| <span class="hljs-comment"># Now in base model's forward we need the decorators and `Unpack` `kwargs`</span> | |
| <span class="hljs-keyword">class</span> <span class="hljs-title class_">MyModel</span>(<span class="hljs-title class_ inherited__">MyPreTrainedModel</span>): | |
| <span class="hljs-meta"> @merge_with_config_defaults </span><span class="hljs-comment"># ← resolves use_cache</span> | |
| <span class="hljs-meta"> @capture_outputs </span><span class="hljs-comment"># ← handles output collection + return_dict</span> | |
| <span class="hljs-keyword">def</span> <span class="hljs-title function_">forward</span>(<span class="hljs-params"> | |
| self, | |
| input_ids: torch.LongTensor | <span class="hljs-literal">None</span> = <span class="hljs-literal">None</span>, | |
| past_key_values: Cache = <span class="hljs-literal">None</span>, | |
| **kwargs: Unpack[TransformersKwargs], | |
| </span>) -> BaseModelOutputWithPast: | |
| <span class="hljs-comment"># No manual collection needed. Just run your layers normally.</span> | |
| hidden_states = self.embed_tokens(input_ids) | |
| <span class="hljs-keyword">for</span> layer <span class="hljs-keyword">in</span> self.layers: | |
| hidden_states = layer(hidden_states, **kwargs) | |
| <span class="hljs-comment"># Return the primary outputs. The decorator will fill in</span> | |
| <span class="hljs-comment"># hidden_states/attentions/cross_attentions automatically.</span> | |
| <span class="hljs-keyword">return</span> BaseModelOutputWithPast( | |
| last_hidden_state=hidden_states, | |
| past_key_values=past_key_values | |
| )<!-- HTML_TAG_END --></pre></div> <h2 class="relative group"><a id="patch-layer-classes" 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="#patch-layer-classes"><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>Patch layer classes</span></h2> <p data-svelte-h="svelte-lkeswn">Output tracing depends on <code>_can_record_outputs</code> pointing at the <em>exact</em> classes a model’s layers instantiate. If you swap a layer implementation for a custom attention kernel, a quantized expert layer, or an architecture variant, those pointers must stay in sync. The <a href="./monkey_patching">patching API</a> provides a clean global registry for keeping <code>_can_record_outputs</code> consistent.</p> <p data-svelte-h="svelte-dp8b4h"><code>register_patch_mapping</code> maps original class names to replacement <code>nn.Module</code> subclasses. Keys can be exact class names or regex patterns. Exact matches take priority. Patterns are tested with <code>re.search()</code>, so unanchored patterns match anywhere in the class name. Registering the same key twice raises <code>ValueError</code> unless you pass <code>overwrite=True</code>.</p> <p data-svelte-h="svelte-mx6c5d">Remove entries with <code>unregister_patch_mapping</code>.</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="language-python "><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> transformers.monkey_patching <span class="hljs-keyword">import</span> register_patch_mapping, unregister_patch_mapping | |
| <span class="hljs-comment"># Exact name – replaces only Qwen2MoeExperts</span> | |
| register_patch_mapping({<span class="hljs-string">"Qwen2MoeExperts"</span>: SequentialExperts}) | |
| <span class="hljs-comment"># Regex – replaces every class whose name ends in "Attention"</span> | |
| register_patch_mapping({<span class="hljs-string">".*Attention$"</span>: FusedAttention}) | |
| <span class="hljs-comment"># Anchored version – only matches Llama2Attention, Llama3Attention, …</span> | |
| register_patch_mapping({<span class="hljs-string">"^Llama\\d+Attention$"</span>: CustomLlamaAttention}) | |
| <span class="hljs-comment"># Same way, custom keys can be removed from the registry by passing the name that was registered</span> | |
| unregister_patch_mapping([<span class="hljs-string">"Qwen2MoeExperts"</span>, <span class="hljs-string">".*Attention$"</span>])<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-zegmky">Once mappings are registered, <code>patch_output_recorders</code> walks every submodule and updates each <code>OutputRecorder.target_class</code> to the registered replacement.</p> <blockquote class="tip" data-svelte-h="svelte-5dfmpi"><p>The <a href="/docs/transformers/pr_36895/en/main_classes/model#transformers.PreTrainedModel.from_pretrained">from_pretrained()</a> method calls <code>patch_output_recorders</code> automatically. You only need to call it yourself when constructing a model directly.</p></blockquote> <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="language-python "><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> transformers.monkey_patching <span class="hljs-keyword">import</span> patch_output_recorders | |
| <span class="hljs-comment"># Built manually, outside from_pretrained</span> | |
| model = Qwen2MoeModel(config) | |
| <span class="hljs-comment"># Without this, _can_record_outputs still points at the original Qwen2MoeExperts class</span> | |
| <span class="hljs-comment"># and hooks will never fire on CustomExperts instances.</span> | |
| patch_output_recorders(model)<!-- 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/model_output_tracing.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_1s77840 = { | |
| assets: "/docs/transformers/pr_36895/en", | |
| base: "/docs/transformers/pr_36895/en", | |
| env: {} | |
| }; | |
| const element = document.currentScript.parentElement; | |
| const data = [null,null]; | |
| Promise.all([ | |
| import("/docs/transformers/pr_36895/en/_app/immutable/entry/start.b88698fb.js"), | |
| import("/docs/transformers/pr_36895/en/_app/immutable/entry/app.ac300a2c.js") | |
| ]).then(([kit, app]) => { | |
| kit.start(app, element, { | |
| node_ids: [0, 579], | |
| data, | |
| form: null, | |
| error: null | |
| }); | |
| }); | |
| } | |
| </script> | |
Xet Storage Details
- Size:
- 25.7 kB
- Xet hash:
- 33d1941fe298d314953bbdb4f185acc40659a3e594968a2d153241fcf1e3b24c
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.