Buckets:
| <meta charset="utf-8" /><meta name="hf:doc:metadata" content="{"title":"Add vision processing components","local":"add-vision-processing-components","sections":[{"title":"Image processors","local":"image-processors","sections":[{"title":"torchvision","local":"torchvision","sections":[],"depth":3},{"title":"PIL","local":"pil","sections":[],"depth":3}],"depth":2},{"title":"Video processor","local":"video-processor","sections":[],"depth":2},{"title":"Register the classes","local":"register-the-classes","sections":[],"depth":2},{"title":"Testing","local":"testing","sections":[{"title":"Image processor tests","local":"image-processor-tests","sections":[],"depth":3},{"title":"Video processor tests","local":"video-processor-tests","sections":[],"depth":3}],"depth":2},{"title":"Next steps","local":"next-steps","sections":[],"depth":2}],"depth":1}"> | |
| <link href="/docs/transformers/main/en/_app/immutable/assets/0.e3b0c442.css" rel="modulepreload"> | |
| <link rel="modulepreload" href="/docs/transformers/main/en/_app/immutable/entry/start.3d6cca8a.js"> | |
| <link rel="modulepreload" href="/docs/transformers/main/en/_app/immutable/chunks/scheduler.31fdf58d.js"> | |
| <link rel="modulepreload" href="/docs/transformers/main/en/_app/immutable/chunks/singletons.6af0ff6e.js"> | |
| <link rel="modulepreload" href="/docs/transformers/main/en/_app/immutable/chunks/index.252883d5.js"> | |
| <link rel="modulepreload" href="/docs/transformers/main/en/_app/immutable/chunks/paths.299a376b.js"> | |
| <link rel="modulepreload" href="/docs/transformers/main/en/_app/immutable/entry/app.b6ccab0d.js"> | |
| <link rel="modulepreload" href="/docs/transformers/main/en/_app/immutable/chunks/preload-helper.c438fa0a.js"> | |
| <link rel="modulepreload" href="/docs/transformers/main/en/_app/immutable/chunks/index.2f76fdf0.js"> | |
| <link rel="modulepreload" href="/docs/transformers/main/en/_app/immutable/nodes/0.f2629ed0.js"> | |
| <link rel="modulepreload" href="/docs/transformers/main/en/_app/immutable/chunks/each.e59479a4.js"> | |
| <link rel="modulepreload" href="/docs/transformers/main/en/_app/immutable/nodes/7.fd7d67df.js"> | |
| <link rel="modulepreload" href="/docs/transformers/main/en/_app/immutable/chunks/CopyLLMTxtMenu.ad38f6ea.js"> | |
| <link rel="modulepreload" href="/docs/transformers/main/en/_app/immutable/chunks/MermaidChart.svelte_svelte_type_style_lang.fd2f7a8a.js"> | |
| <link rel="modulepreload" href="/docs/transformers/main/en/_app/immutable/chunks/IconCopy.ac192424.js"> | |
| <link rel="modulepreload" href="/docs/transformers/main/en/_app/immutable/chunks/CodeBlock.e52df5d6.js"><!-- HEAD_svelte-u9bgzb_START --><meta name="hf:doc:metadata" content="{"title":"Add vision processing components","local":"add-vision-processing-components","sections":[{"title":"Image processors","local":"image-processors","sections":[{"title":"torchvision","local":"torchvision","sections":[],"depth":3},{"title":"PIL","local":"pil","sections":[],"depth":3}],"depth":2},{"title":"Video processor","local":"video-processor","sections":[],"depth":2},{"title":"Register the classes","local":"register-the-classes","sections":[],"depth":2},{"title":"Testing","local":"testing","sections":[{"title":"Image processor tests","local":"image-processor-tests","sections":[],"depth":3},{"title":"Video processor tests","local":"video-processor-tests","sections":[],"depth":3}],"depth":2},{"title":"Next steps","local":"next-steps","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="add-vision-processing-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="#add-vision-processing-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>Add vision processing components</span></h1> <p data-svelte-h="svelte-i1rbh">Adding a vision model requires image or video processing components on top of the standard <a href="./modular_transformers">modular</a> approach. Image-only models need image processors and video models need a video processor, both of which are accessible behind the <a href="/docs/transformers/main/en/model_doc/auto#transformers.AutoImageProcessor">AutoImageProcessor</a> and <a href="/docs/transformers/main/en/model_doc/auto#transformers.AutoVideoProcessor">AutoVideoProcessor</a> entry points.</p> <blockquote class="note" data-svelte-h="svelte-5giosi"><p>For the modeling and config steps, follow the <a href="./modular_transformers">modular</a> guide first.</p></blockquote> <h2 class="relative group"><a id="image-processors" 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="#image-processors"><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>Image processors</span></h2> <p data-svelte-h="svelte-1jt4hsn">Create image processors when the model consumes images. The <a href="https://docs.pytorch.org/vision/stable/index.html" rel="nofollow">torchvision</a> backend is the default and supports GPU acceleration. <a href="https://pillow.readthedocs.io/en/stable/index.html" rel="nofollow">PIL</a> is the fallback when torchvision isn’t available.</p> <p data-svelte-h="svelte-1vy18ci">Both image processor classes share the same preprocessing logic but have different backends. Their constructor signatures and default values must be identical. <a href="/docs/transformers/main/en/model_doc/auto#transformers.AutoImageProcessor.from_pretrained">AutoImageProcessor.from_pretrained()</a> selects the backend at load time and falls back to PIL when torchvision isn’t available. Mismatched signatures cause the same saved config to behave differently across environments.</p> <h3 class="relative group"><a id="torchvision" 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="#torchvision"><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>torchvision</span></h3> <p data-svelte-h="svelte-1a777xy">Create <code>image_processing_<model_name>.py</code> with a class that inherits from <a href="/docs/transformers/main/en/main_classes/image_processor#transformers.TorchvisionBackend">TorchvisionBackend</a>. If your processor needs custom parameters beyond the standard [ImagesKwargs], define a kwargs class.</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-py "><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> ...image_processing_backends <span class="hljs-keyword">import</span> TorchvisionBackend | |
| <span class="hljs-keyword">from</span> ...image_utils <span class="hljs-keyword">import</span> OPENAI_CLIP_MEAN, OPENAI_CLIP_STD, PILImageResampling | |
| <span class="hljs-keyword">from</span> ...processing_utils <span class="hljs-keyword">import</span> ImagesKwargs, Unpack | |
| <span class="hljs-keyword">from</span> ...utils <span class="hljs-keyword">import</span> auto_docstring | |
| <span class="hljs-keyword">class</span> <span class="hljs-title class_">MyModelImageProcessorKwargs</span>(ImagesKwargs, total=<span class="hljs-literal">False</span>): | |
| tile_size: <span class="hljs-built_in">int</span> <span class="hljs-comment"># any model-specific kwargs</span> | |
| <span class="hljs-meta">@auto_docstring</span> | |
| <span class="hljs-keyword">class</span> <span class="hljs-title class_">MyModelImageProcessor</span>(<span class="hljs-title class_ inherited__">TorchvisionBackend</span>): | |
| resample = PILImageResampling.BICUBIC | |
| image_mean = OPENAI_CLIP_MEAN | |
| image_std = OPENAI_CLIP_STD | |
| size = {<span class="hljs-string">"shortest_edge"</span>: <span class="hljs-number">224</span>} | |
| do_resize = <span class="hljs-literal">True</span> | |
| do_rescale = <span class="hljs-literal">True</span> | |
| do_normalize = <span class="hljs-literal">True</span> | |
| do_convert_rgb = <span class="hljs-literal">True</span> | |
| <span class="hljs-keyword">def</span> <span class="hljs-title function_">__init__</span>(<span class="hljs-params">self, **kwargs: Unpack[MyModelImageProcessorKwargs]</span>): | |
| <span class="hljs-built_in">super</span>().__init__(**kwargs)<!-- HTML_TAG_END --></pre></div> <blockquote class="tip" data-svelte-h="svelte-14rwt12"><p>See <a href="/docs/transformers/main/en/model_doc/llava_onevision#transformers.LlavaOnevisionImageProcessor">LlavaOnevisionImageProcessor</a> for reference.</p></blockquote> <h3 class="relative group"><a id="pil" 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="#pil"><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>PIL</span></h3> <p data-svelte-h="svelte-1if2mef">Create <code>image_processing_pil_<model_name>.py</code> with a class that inherits from <a href="/docs/transformers/main/en/main_classes/image_processor#transformers.PilBackend">PilBackend</a>. Duplicate the kwargs class here instead of importing it from the torchvision file because it can fail when torchvision isn’t installed. Add an <code># Adapted from</code> comment so the two stay in sync. For processors with no custom parameters, use <a href="/docs/transformers/main/en/main_classes/processors#transformers.ImagesKwargs">ImagesKwargs</a> directly.</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-py "><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> ...image_processing_backends <span class="hljs-keyword">import</span> PilBackend | |
| <span class="hljs-keyword">from</span> ...image_utils <span class="hljs-keyword">import</span> OPENAI_CLIP_MEAN, OPENAI_CLIP_STD, PILImageResampling | |
| <span class="hljs-keyword">from</span> ...processing_utils <span class="hljs-keyword">import</span> ImagesKwargs, Unpack | |
| <span class="hljs-keyword">from</span> ...utils <span class="hljs-keyword">import</span> auto_docstring | |
| <span class="hljs-comment"># Adapted from transformers.models.my_model.image_processing_my_model.MyModelImageProcessorKwargs</span> | |
| <span class="hljs-keyword">class</span> <span class="hljs-title class_">MyModelImageProcessorKwargs</span>(ImagesKwargs, total=<span class="hljs-literal">False</span>): | |
| tile_size: <span class="hljs-built_in">int</span> <span class="hljs-comment"># any model-specific kwargs</span> | |
| <span class="hljs-meta">@auto_docstring</span> | |
| <span class="hljs-keyword">class</span> <span class="hljs-title class_">MyModelImageProcessorPil</span>(<span class="hljs-title class_ inherited__">PilBackend</span>): | |
| resample = PILImageResampling.BICUBIC | |
| image_mean = OPENAI_CLIP_MEAN | |
| image_std = OPENAI_CLIP_STD | |
| size = {<span class="hljs-string">"shortest_edge"</span>: <span class="hljs-number">224</span>} | |
| do_resize = <span class="hljs-literal">True</span> | |
| do_rescale = <span class="hljs-literal">True</span> | |
| do_normalize = <span class="hljs-literal">True</span> | |
| do_convert_rgb = <span class="hljs-literal">True</span> | |
| <span class="hljs-keyword">def</span> <span class="hljs-title function_">__init__</span>(<span class="hljs-params">self, **kwargs: Unpack[MyModelImageProcessorKwargs]</span>): | |
| <span class="hljs-built_in">super</span>().__init__(**kwargs)<!-- HTML_TAG_END --></pre></div> <blockquote class="tip" data-svelte-h="svelte-4imp8k"><p>See <a href="/docs/transformers/main/en/model_doc/llava_onevision#transformers.LlavaOnevisionImageProcessorPil">LlavaOnevisionImageProcessorPil</a> for reference.</p></blockquote> <h2 class="relative group"><a id="video-processor" 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="#video-processor"><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>Video processor</span></h2> <p data-svelte-h="svelte-9qx83a">Add a video processor when the model consumes videos or sampled video frames.</p> <p data-svelte-h="svelte-1a7uslx">Create <code>video_processing_<model_name>.py</code> in the model directory. <a href="/docs/transformers/main/en/main_classes/video_processor#transformers.BaseVideoProcessor">BaseVideoProcessor</a> inherits from the <a href="/docs/transformers/main/en/main_classes/image_processor#transformers.TorchvisionBackend">TorchvisionBackend</a> and provides shared decoding, frame sampling, resizing, rescaling, normalization, saving, and loading behavior.</p> <p data-svelte-h="svelte-13oeg4m">The class attributes are the default preprocessing values. Users can override them at initialization or call time. Use the same names as <a href="/docs/transformers/main/en/main_classes/processors#transformers.VideosKwargs">VideosKwargs</a> when possible, such as <code>size</code>, <code>crop_size</code>, <code>do_resize</code>, <code>do_sample_frames</code>, <code>num_frames</code>, and <code>fps</code>.</p> <p data-svelte-h="svelte-1ufprbj">Define a kwargs class if your video processor needs custom parameters beyond the standard <a href="/docs/transformers/main/en/main_classes/processors#transformers.VideosKwargs">VideosKwargs</a>. Set it as <code>valid_kwargs</code> and use it to annotate <code>__init__</code> for both runtime validation and the auto-generated docstring.</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-py "><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> ...processing_utils <span class="hljs-keyword">import</span> Unpack, VideosKwargs | |
| <span class="hljs-keyword">from</span> ...utils <span class="hljs-keyword">import</span> auto_docstring | |
| <span class="hljs-keyword">from</span> ...video_processing_utils <span class="hljs-keyword">import</span> BaseVideoProcessor | |
| <span class="hljs-keyword">class</span> <span class="hljs-title class_">MyModelVideoProcessorKwargs</span>(VideosKwargs, total=<span class="hljs-literal">False</span>): | |
| min_frames: <span class="hljs-built_in">int</span> | |
| max_frames: <span class="hljs-built_in">int</span> | |
| <span class="hljs-meta">@auto_docstring</span> | |
| <span class="hljs-keyword">class</span> <span class="hljs-title class_">MyModelVideoProcessor</span>(<span class="hljs-title class_ inherited__">BaseVideoProcessor</span>): | |
| size = {<span class="hljs-string">"shortest_edge"</span>: <span class="hljs-number">224</span>} | |
| crop_size = {<span class="hljs-string">"height"</span>: <span class="hljs-number">224</span>, <span class="hljs-string">"width"</span>: <span class="hljs-number">224</span>} | |
| do_resize = <span class="hljs-literal">True</span> | |
| do_center_crop = <span class="hljs-literal">True</span> | |
| do_normalize = <span class="hljs-literal">True</span> | |
| do_sample_frames = <span class="hljs-literal">True</span> | |
| num_frames = <span class="hljs-number">16</span> | |
| model_input_names = [<span class="hljs-string">"pixel_values_videos"</span>] | |
| valid_kwargs = MyModelVideoProcessorKwargs | |
| <span class="hljs-keyword">def</span> <span class="hljs-title function_">__init__</span>(<span class="hljs-params">self, **kwargs: Unpack[MyModelVideoProcessorKwargs]</span>): | |
| <span class="hljs-built_in">super</span>().__init__(**kwargs)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1dtcwr4">Override <a href="/docs/transformers/main/en/main_classes/video_processor#transformers.BaseVideoProcessor.sample_frames">sample_frames()</a> only when the model requires a sampling rule that the base uniform sampler can’t express. For example, some models enforce a minimum or maximum number of frames, or sample based on model-specific constraints.</p> <p data-svelte-h="svelte-360sbo">If the model’s forward method expects a legacy input name, override <code>preprocess</code> and rename the key after calling the base implementation.</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-py "><!-- HTML_TAG_START --><span class="hljs-keyword">class</span> <span class="hljs-title class_">MyModelVideoProcessor</span>(<span class="hljs-title class_ inherited__">BaseVideoProcessor</span>): | |
| model_input_names = [<span class="hljs-string">"pixel_values"</span>] | |
| <span class="hljs-keyword">def</span> <span class="hljs-title function_">preprocess</span>(<span class="hljs-params">self, videos, **kwargs</span>): | |
| batch = <span class="hljs-built_in">super</span>().preprocess(videos, **kwargs) | |
| batch[<span class="hljs-string">"pixel_values"</span>] = batch.pop(<span class="hljs-string">"pixel_values_videos"</span>) | |
| <span class="hljs-keyword">return</span> batch<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-3esgu1">Save the video processor with the checkpoint by instantiating it in the conversion script and calling <a href="/docs/transformers/main/en/main_classes/video_processor#transformers.BaseVideoProcessor.save_pretrained">save_pretrained()</a>. If a <a href="/docs/transformers/main/en/main_classes/processors#transformers.ProcessorMixin">ProcessorMixin</a> wraps the video processor, call <a href="/docs/transformers/main/en/model_doc/wav2vec2#transformers.Wav2Vec2Processor.save_pretrained">save_pretrained()</a> instead. Do not manually create or edit preprocessing config files.</p> <blockquote class="tip" data-svelte-h="svelte-17hqcdq"><p>See <a href="/docs/transformers/main/en/model_doc/qwen3_vl#transformers.Qwen3VLVideoProcessor">Qwen3VLVideoProcessor</a> for reference.</p></blockquote> <h2 class="relative group"><a id="register-the-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="#register-the-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>Register the classes</span></h2> <p data-svelte-h="svelte-7c6wc9">Expose the processing classes from the model package <code>__init__.py</code>. Follow the lazy import pattern used by nearby models and guard imports with the same optional dependencies required by each backend.</p> <p data-svelte-h="svelte-18lvmsm">Map the new classes to the model config so the <code>Auto</code> classes can load them. The generated auto mapping file has a warning at the top. Do not edit it by hand. Add or update the model config, then run:</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-bash "><!-- HTML_TAG_START -->python utils/check_auto.py --fix_and_overwrite<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1c3w5ey">After the mapping is generated, verify the model type appears in the relevant mappings in <code>src/transformers/models/auto/auto_mappings.py</code>.</p> <ul data-svelte-h="svelte-13399fm"><li><code>IMAGE_PROCESSOR_MAPPING_NAMES</code> for <a href="/docs/transformers/main/en/model_doc/auto#transformers.AutoImageProcessor">AutoImageProcessor</a></li> <li><code>VIDEO_PROCESSOR_MAPPING_NAMES</code> for <a href="/docs/transformers/main/en/model_doc/auto#transformers.AutoVideoProcessor">AutoVideoProcessor</a></li></ul> <h2 class="relative group"><a id="testing" 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="#testing"><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>Testing</span></h2> <p data-svelte-h="svelte-1wxd6no">Add tests for each vision processing component in the model test directory. Image and video processor tests follow the same pattern. Inherit from the shared mixin, indicate the fast and slow processing classes when automatic discovery isn’t enough, provide model-specific init kwargs, and override the input name when the model uses a non-default output key.</p> <h3 class="relative group"><a id="image-processor-tests" 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="#image-processor-tests"><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>Image processor tests</span></h3> <p data-svelte-h="svelte-68f8g5">Image processor tests usually live in <code>tests/models/<model_name>/test_image_processing_<model_name>.py</code> and inherit from <code>ImageProcessingTestMixin</code>.</p> <p data-svelte-h="svelte-xv7dsz">The image processing mixin finds the image processor classes from <code>IMAGE_PROCESSOR_MAPPING_NAMES</code>. Expose model-specific defaults through <code>image_processor_dict</code>. Add a tester object only when you need reusable dummy inputs or helper methods for focused tests.</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-py "><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> transformers.testing_utils <span class="hljs-keyword">import</span> require_torch, require_vision | |
| <span class="hljs-keyword">from</span> ...test_image_processing_common <span class="hljs-keyword">import</span> ImageProcessingTestMixin | |
| <span class="hljs-meta">@require_torch</span> | |
| <span class="hljs-meta">@require_vision</span> | |
| <span class="hljs-keyword">class</span> <span class="hljs-title class_">MyModelImageProcessingTest</span>(ImageProcessingTestMixin, unittest.TestCase): | |
| <span class="hljs-meta"> @property</span> | |
| <span class="hljs-keyword">def</span> <span class="hljs-title function_">image_processor_dict</span>(<span class="hljs-params">self</span>): | |
| <span class="hljs-keyword">return</span> {<span class="hljs-string">"size"</span>: {<span class="hljs-string">"shortest_edge"</span>: <span class="hljs-number">224</span>}, <span class="hljs-string">"do_resize"</span>: <span class="hljs-literal">True</span>}<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1m5x7q5">Add focused tests for behavior the mixin can’t infer, such as custom resizing rules or model-specific kwargs.</p> <h3 class="relative group"><a id="video-processor-tests" 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="#video-processor-tests"><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>Video processor tests</span></h3> <p data-svelte-h="svelte-134qk85">Video processor tests usually live in <code>tests/models/<model_name>/test_video_processing_<model_name>.py</code> and inherit from <code>VideoProcessingTestMixin</code>. Set <code>fast_video_processing_class</code>, define <code>video_processor_dict</code>, and override <code>input_name</code> if the model uses a key other than <code>pixel_values_videos</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-py "><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> transformers.testing_utils <span class="hljs-keyword">import</span> require_torch, require_vision | |
| <span class="hljs-keyword">from</span> transformers.utils <span class="hljs-keyword">import</span> is_torchvision_available | |
| <span class="hljs-keyword">from</span> ...test_video_processing_common <span class="hljs-keyword">import</span> VideoProcessingTestMixin | |
| <span class="hljs-meta">@require_torch</span> | |
| <span class="hljs-meta">@require_vision</span> | |
| <span class="hljs-keyword">class</span> <span class="hljs-title class_">MyModelVideoProcessingTest</span>(VideoProcessingTestMixin, unittest.TestCase): | |
| fast_video_processing_class = MyModelVideoProcessor <span class="hljs-keyword">if</span> is_torchvision_available() <span class="hljs-keyword">else</span> <span class="hljs-literal">None</span> | |
| input_name = <span class="hljs-string">"pixel_values_videos"</span> | |
| <span class="hljs-meta"> @property</span> | |
| <span class="hljs-keyword">def</span> <span class="hljs-title function_">video_processor_dict</span>(<span class="hljs-params">self</span>): | |
| <span class="hljs-keyword">return</span> {<span class="hljs-string">"size"</span>: {<span class="hljs-string">"shortest_edge"</span>: <span class="hljs-number">224</span>}, <span class="hljs-string">"num_frames"</span>: <span class="hljs-number">16</span>}<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-8by4sn">Add focused video tests for frame sampling, metadata handling, decoded video inputs, list-of-frame inputs, and output shapes. If your processor renames <code>pixel_values_videos</code>, assert the renamed key is returned.</p> <p data-svelte-h="svelte-59qjmz">If the model also has a <a href="/docs/transformers/main/en/main_classes/processors#transformers.ProcessorMixin">ProcessorMixin</a> that wraps the image or video processor, add <code>tests/models/<model_name>/test_processing_<model_name>.py</code> and inherit from <code>ProcessorTesterMixin</code>. Set <code>processor_class</code> and override <code>_setup_<component>()</code> class methods for components that can’t be constructed without arguments. Use <code>_setup_test_attributes()</code> to expose placeholder tokens used by the common processor tests.</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-py "><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> ...test_processing_common <span class="hljs-keyword">import</span> ProcessorTesterMixin | |
| <span class="hljs-keyword">class</span> <span class="hljs-title class_">MyModelProcessorTest</span>(ProcessorTesterMixin, unittest.TestCase): | |
| processor_class = MyModelProcessor | |
| <span class="hljs-meta"> @classmethod</span> | |
| <span class="hljs-keyword">def</span> <span class="hljs-title function_">_setup_image_processor</span>(<span class="hljs-params">cls</span>): | |
| <span class="hljs-keyword">return</span> cls._get_component_class_from_processor(<span class="hljs-string">"image_processor"</span>)(size={<span class="hljs-string">"shortest_edge"</span>: <span class="hljs-number">224</span>}) | |
| <span class="hljs-meta"> @classmethod</span> | |
| <span class="hljs-keyword">def</span> <span class="hljs-title function_">_setup_video_processor</span>(<span class="hljs-params">cls</span>): | |
| <span class="hljs-keyword">return</span> cls._get_component_class_from_processor(<span class="hljs-string">"video_processor"</span>)(num_frames=<span class="hljs-number">2</span>) | |
| <span class="hljs-meta"> @classmethod</span> | |
| <span class="hljs-keyword">def</span> <span class="hljs-title function_">_setup_test_attributes</span>(<span class="hljs-params">cls, processor</span>): | |
| cls.image_token = <span class="hljs-built_in">getattr</span>(processor, <span class="hljs-string">"image_token"</span>, <span class="hljs-string">""</span>) | |
| cls.video_token = <span class="hljs-built_in">getattr</span>(processor, <span class="hljs-string">"video_token"</span>, <span class="hljs-string">""</span>)<!-- HTML_TAG_END --></pre></div> <h2 class="relative group"><a id="next-steps" 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="#next-steps"><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>Next steps</span></h2> <ul data-svelte-h="svelte-8w33b4"><li>Read the <a href="./auto_docstring">Auto-generating docstrings</a> guide to auto-generate consistent docstrings with <code>@auto_docstring</code>.</li> <li>Read the <a href="./image_processors">Image processors</a> and <a href="./video_processors">Video processors</a> guides for user-facing preprocessing behavior.</li></ul> <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/add_vision_processing_components.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_rv114u = { | |
| assets: "/docs/transformers/main/en", | |
| base: "/docs/transformers/main/en", | |
| env: {} | |
| }; | |
| const element = document.currentScript.parentElement; | |
| const data = [null,null]; | |
| Promise.all([ | |
| import("/docs/transformers/main/en/_app/immutable/entry/start.3d6cca8a.js"), | |
| import("/docs/transformers/main/en/_app/immutable/entry/app.b6ccab0d.js") | |
| ]).then(([kit, app]) => { | |
| kit.start(app, element, { | |
| node_ids: [0, 7], | |
| data, | |
| form: null, | |
| error: null | |
| }); | |
| }); | |
| } | |
| </script> | |
Xet Storage Details
- Size:
- 48.4 kB
- Xet hash:
- e708450b5bfaac54a24f2fa977f1a78079d6d076d6b6cd2293218ee3a450a7b8
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.