Buckets:
| <meta charset="utf-8" /><meta name="hf:doc:metadata" content="{"title":"FLUX 3 Image Klein","local":"flux-3-image-klein","sections":[{"title":"提示词上采样","local":"prompt-upsampling","sections":[],"depth":2},{"title":"图像到图像","local":"image-to-image","sections":[],"depth":2},{"title":"布局控制","local":"layout-control","sections":[],"depth":2},{"title":"引导比例","local":"guidance-scale","sections":[],"depth":2},{"title":"步数蒸馏模型","local":"step-distilled-model","sections":[],"depth":2},{"title":"可复现的生成","local":"reproducible-generation","sections":[],"depth":2},{"title":"工作流程","local":"workflows","sections":[],"depth":2},{"title":"逐步回调","local":"per-step-callbacks","sections":[],"depth":2},{"title":"内存","local":"memory","sections":[],"depth":2},{"title":"单文件加载","local":"single-file-loading","sections":[],"depth":2},{"title":"Flux3ModularPipeline","local":"diffusers.Flux3ModularPipeline","sections":[],"depth":2},{"title":"Flux3AutoBlocks","local":"diffusers.Flux3AutoBlocks","sections":[],"depth":2},{"title":"Flux3DistilledAutoBlocks","local":"diffusers.Flux3DistilledAutoBlocks","sections":[],"depth":2},{"title":"Flux3DistilledModularPipeline","local":"diffusers.Flux3DistilledModularPipeline","sections":[],"depth":2}],"depth":1}"/> | |
| <link href="/docs/diffusers/v0.43.0/en/_app/immutable/entry/start.1gaXe03k.js" rel="modulepreload"> | |
| <link href="/docs/diffusers/v0.43.0/en/_app/immutable/chunks/BObs6oQO.js" rel="modulepreload"> | |
| <link href="/docs/diffusers/v0.43.0/en/_app/immutable/chunks/DPnyVmXs.js" rel="modulepreload"> | |
| <link href="/docs/diffusers/v0.43.0/en/_app/immutable/entry/app.BwdYRbCn.js" rel="modulepreload"> | |
| <link href="/docs/diffusers/v0.43.0/en/_app/immutable/chunks/DSGUT_OW.js" rel="modulepreload"> | |
| <link href="/docs/diffusers/v0.43.0/en/_app/immutable/chunks/C045MWWx.js" rel="modulepreload"> | |
| <link href="/docs/diffusers/v0.43.0/en/_app/immutable/chunks/6tuJ3nyi.js" rel="modulepreload"> | |
| <link href="/docs/diffusers/v0.43.0/en/_app/immutable/chunks/DsnmJJEf.js" rel="modulepreload"> | |
| <link href="/docs/diffusers/v0.43.0/en/_app/immutable/nodes/0.CornNeQR.js" rel="modulepreload"> | |
| <link href="/docs/diffusers/v0.43.0/en/_app/immutable/chunks/IaDvMxLQ.js" rel="modulepreload"> | |
| <link href="/docs/diffusers/v0.43.0/en/_app/immutable/nodes/180.DpDO5C2D.js" rel="modulepreload"> | |
| <link href="/docs/diffusers/v0.43.0/en/_app/immutable/chunks/bKdwVYpU.js" rel="modulepreload"> | |
| <!--hchutw--><meta name="hf:doc:metadata" content="{"title":"FLUX 3 Image Klein","local":"flux-3-image-klein","sections":[{"title":"Prompt upsampling","local":"prompt-upsampling","sections":[],"depth":2},{"title":"Image-to-image","local":"image-to-image","sections":[],"depth":2},{"title":"Layout control","local":"layout-control","sections":[],"depth":2},{"title":"Guidance scale","local":"guidance-scale","sections":[],"depth":2},{"title":"Step-distilled model","local":"step-distilled-model","sections":[],"depth":2},{"title":"Reproducible generation","local":"reproducible-generation","sections":[],"depth":2},{"title":"Workflows","local":"workflows","sections":[],"depth":2},{"title":"Per-step callbacks","local":"per-step-callbacks","sections":[],"depth":2},{"title":"Memory","local":"memory","sections":[],"depth":2},{"title":"Single-file loading","local":"single-file-loading","sections":[],"depth":2},{"title":"Flux3ModularPipeline","local":"diffusers.Flux3ModularPipeline","sections":[],"depth":2},{"title":"Flux3AutoBlocks","local":"diffusers.Flux3AutoBlocks","sections":[],"depth":2},{"title":"Flux3DistilledAutoBlocks","local":"diffusers.Flux3DistilledAutoBlocks","sections":[],"depth":2},{"title":"Flux3DistilledModularPipeline","local":"diffusers.Flux3DistilledModularPipeline","sections":[],"depth":2}],"depth":1}"/><!----> | |
| <link href="/docs/diffusers/v0.43.0/en/_app/immutable/assets/0.tn0RQdqM.css" rel="modulepreload"> <!--[--><!--[0--><!--[--><!--[0--><!--[--><p></p> <!--[0--><h1 class="relative group"><a id="flux-3-image-klein" 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="#flux-3-image-klein"><span><svg 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><!---->FLUX 3 Image Klein<!----></span></h1><!--]--><!----> <p>FLUX 3 Image Klein 是 Black Forest Labs 推出的 5.6B 图像生成模型:单个 Transformer(<a href="/docs/diffusers/v0.43.0/zh/api/models/flux3_transformer#diffusers.Flux3Transformer2DModel">Flux3Transformer2DModel</a>)可根据文本提示生成图像,也可结合一张或多张参考图像进行生成。文本条件来自 Qwen3-VL,图像则使用 FLUX.2 打包自编码器(<code>AutoencoderKLFlux2</code>)。</p> <p>FLUX 3 Image Klein 以<a href="../../modular_diffusers/overview">模块化流水线</a>的形式发布。从文本到图像生成开始:</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 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> 已复制</div><!----></button><!----></div> <pre class="language-python "><!----><span class="hljs-keyword">import</span> torch | |
| <span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> ModularPipeline | |
| pipe = ModularPipeline.from_pretrained(<span class="hljs-string">"black-forest-labs/flux-3-image-klein-base"</span>) | |
| pipe.load_components(dtype={<span class="hljs-string">"vae"</span>: torch.float32, <span class="hljs-string">"default"</span>: torch.bfloat16}) | |
| pipe.to(<span class="hljs-string">"cuda"</span>) | |
| image = pipe( | |
| prompt=<span class="hljs-string">"A red fox in a snow-covered forest clearing at golden hour"</span>, | |
| generator=torch.Generator().manual_seed(<span class="hljs-number">0</span>), | |
| output=<span class="hljs-string">"images"</span>, | |
| )[<span class="hljs-number">0</span>] | |
| image.save(<span class="hljs-string">"fox.png"</span>)<!----></pre></div><!----> <p><code>prompt</code> 是唯一必需的输入:该流水线会携带推荐的采样设置(50 步、guidance scale 3.5、scheduler shift 6.93),并默认生成 1024 x 1024 的图像。每次调用仅接受一个 prompt;若需为同一 prompt 生成多张图像,请使用 <code>num_images_per_prompt</code>。</p> <p>如上例所示,加载和运行方式与 <code>DiffusionPipeline</code> 略有不同。加载是惰性的:<a href="/docs/diffusers/v0.43.0/zh/api/modular_diffusers/pipeline#diffusers.ModularPipeline.from_pretrained">from_pretrained()</a> 仅构建流水线,你需要调用 <a href="/docs/diffusers/v0.43.0/zh/api/modular_diffusers/pipeline#diffusers.ModularPipeline.load_components">load_components()</a> 来加载权重。调用后返回流水线状态,你可以通过 <code>output</code> 来指定所需内容:此处为 <code>output="images"</code>,或者使用 <code>output=["images", "prompt"]</code> 以同时获取提示词上采样器生成的说明文本。</p> <blockquote class="tip"><p>如果您首次使用 <code>ModularPipeline</code>,请先阅读<a href="../../modular_diffusers/quickstart">快速入门</a>;如果您熟悉标准流水线,请阅读<a href="../../modular_diffusers/coming_from_diffusion_pipeline">从 DiffusionPipeline 迁移</a>。除了加载和输出之外,还有几处不同:<a href="#guidance-scale">guidance</a>、<a href="#per-step-callbacks">callbacks</a> 和 <a href="#memory">offloading</a>。</p></blockquote> <!--[1--><h2 class="relative group"><a id="prompt-upsampling" 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="#prompt-upsampling"><span><svg 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><!---->提示词上采样<!----></span></h2><!--]--><!----> <p>提示词至关重要:FLUX 3 Image Klein 需要完整的场景描述,涵盖主体、构图、光照和氛围,而非简短的提示词;直接输入简短请求只会得到泛化的输出。正因如此,该流水线在处理每个请求时(无论是文生图还是图像编辑)都会先运行提示词扩展器。这是 Black Forest Labs 的一个独立 9B 模型(<a href="https://huggingface.co/black-forest-labs/flux-3-image-klein-pup" rel="nofollow">black-forest-labs/flux-3-image-klein-pup</a>),它会在文本编码器之前将 <code>prompt</code> 扩展为上述类型的场景描述,因此只需提供简短的提示词即可。</p> <p>提示词扩展器会将简短的请求(在编辑时还包括参考图像)转化为模型所需的完整场景描述:涵盖中景与风格、场景、光照与镜头,以及每个主体的颜色、材质和在画面中的位置。一个模型即可覆盖文生图、基于参考图像的编辑以及基于框的布局。它生成的描述会作为 <code>prompt</code> 输出返回。在上面的示例中,可以请求同时返回图像,以查看图像模型实际接收到的内容:</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 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> 已复制</div><!----></button><!----></div> <pre class="language-python "><!---->out = pipe( | |
| prompt=<span class="hljs-string">"A red fox in a snow-covered forest clearing at golden hour"</span>, | |
| generator=torch.Generator().manual_seed(<span class="hljs-number">0</span>), | |
| output=[<span class="hljs-string">"images"</span>, <span class="hljs-string">"prompt"</span>], | |
| ) | |
| <span class="hljs-built_in">print</span>(out[<span class="hljs-string">"prompt"</span>])<!----></pre></div><!----> <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 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> 已复制</div><!----></button><!----></div> <pre class=" "><!---->Full-color outdoor wildlife photograph taken <span class="hljs-built_in">from</span> <span class="hljs-keyword">a</span> low eye-level perspective <span class="hljs-keyword">in</span> <span class="hljs-keyword">a</span> snow-covered forest clearing <span class="hljs-keyword">at</span> golden hour. In <span class="hljs-keyword">the</span> center foreground, <span class="hljs-keyword">a</span> red fox stands <span class="hljs-keyword">in</span> profile facing left, its body turned toward <span class="hljs-keyword">the</span> viewer <span class="hljs-keyword">and</span> its head raised <span class="hljs-keyword">as</span> <span class="hljs-keyword">it</span> looks across <span class="hljs-keyword">the</span> clearing. The fox has dense rust-red winter fur, <span class="hljs-keyword">a</span> white throat <span class="hljs-keyword">and</span> chest, <span class="hljs-keyword">black</span> <span class="hljs-built_in">lower</span> legs, pointed ears <span class="hljs-keyword">with</span> dark tips, <span class="hljs-keyword">and</span> <span class="hljs-keyword">a</span> <span class="hljs-keyword">long</span> bushy tail ending <span class="hljs-keyword">in</span> white. ...<!----></pre></div><!----> <p>完整字幕长度约为 240 个词;<code>max_new_tokens</code> 用于限制其长度(默认值为 3584)。</p> <p>您可以打印该步骤的 <code>doc</code>,以查看其输入和输出内容:</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 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> 已复制</div><!----></button><!----></div> <pre class="language-python "><!----><span class="hljs-built_in">print</span>(pipe.blocks.sub_blocks[<span class="hljs-string">"prompt_upsampler"</span>].doc)<!----></pre></div><!----> <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 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> 已复制</div><!----></button><!----></div> <pre class=" "><!----> <span class="hljs-symbol">Inputs</span>: | |
| prompt (<span class="hljs-string">`str`</span>): | |
| <span class="hljs-symbol">The</span> edit instruction to expand into the full caption (one per call). | |
| image (<span class="hljs-string">`Image | list`</span>, *optional*): | |
| <span class="hljs-symbol">The</span> reference image(s) of the edit, at most <span class="hljs-number">10</span>; the same ones the pipeline receives. | |
| boxes (<span class="hljs-string">`list`</span>, *optional*): | |
| <span class="hljs-symbol">Layout</span> boxes, each {<span class="hljs-string">'id'</span>, <span class="hljs-string">'desc'</span>, <span class="hljs-string">'from'</span>, <span class="hljs-string">'src_bbox'</span>, <span class="hljs-string">'tgt_bbox'</span>} with the boxes as [y_min, x_min, y_max, x_max] in | |
| <span class="hljs-number">0</span><span class="hljs-number">-1</span>: ... | |
| max_new_tokens (<span class="hljs-string">`int`</span>, *optional*, defaults to <span class="hljs-number">3584</span>): | |
| <span class="hljs-symbol">The</span> most tokens the upsampler may generate for the caption. | |
| <span class="hljs-symbol">Outputs</span>: | |
| prompt (<span class="hljs-string">`str`</span>): | |
| <span class="hljs-symbol">The</span> full caption, replacing the request.<!----></pre></div><!----> <p>与流水线中的其他块一样,<code>prompt_upsampler</code> 块也是独立运行的。推荐的做法是将其从块中弹出:弹出的块将成为一个独立的提示词扩展器,而剩余的块则构成一个流水线,将扩展后的描述作为 <code>prompt</code> 输入:</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 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> 已复制</div><!----></button><!----></div> <pre class="language-python "><!---->blocks = ModularPipeline.from_pretrained(<span class="hljs-string">"black-forest-labs/flux-3-image-klein-base"</span>).blocks | |
| pup_block = blocks.sub_blocks.pop(<span class="hljs-string">"prompt_upsampler"</span>) | |
| pup = pup_block.init_pipeline(<span class="hljs-string">"black-forest-labs/flux-3-image-klein-base"</span>) | |
| pup.load_components(dtype=torch.bfloat16) | |
| pup.to(<span class="hljs-string">"cuda"</span>) | |
| caption = pup(prompt=<span class="hljs-string">"A red fox in a snow-covered forest clearing at golden hour"</span>, output=<span class="hljs-string">"prompt"</span>) | |
| pipe = blocks.init_pipeline(<span class="hljs-string">"black-forest-labs/flux-3-image-klein-base"</span>) | |
| pipe.load_components(dtype={<span class="hljs-string">"vae"</span>: torch.float32, <span class="hljs-string">"default"</span>: torch.bfloat16}) | |
| pipe.to(<span class="hljs-string">"cuda"</span>) | |
| image = pipe(prompt=caption, generator=torch.Generator().manual_seed(<span class="hljs-number">0</span>), output=<span class="hljs-string">"images"</span>)[<span class="hljs-number">0</span>]<!----></pre></div><!----> <p>这样,它可以作为一个独立的阶段进行部署,运行在单独的 GPU 上或位于独立的端点之后,UI 可以展示其生成的描述文本,并允许用户在生成图像前对其进行编辑。任何其他 LLM 都可以替代上采样器,使用相同的系统提示词:将上采样器仓库中的 <code>system_prompts["t2i"]</code>(或 <code>"i2i"</code>,其中参考图像位于用户消息的开头)作为系统提示词,将请求作为用户消息,并将其返回的描述文本作为 <code>prompt</code> 传递给 <code>pipe</code>。</p> <!--[1--><h2 class="relative group"><a id="image-to-image" 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-to-image"><span><svg 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><!---->图像到图像<!----></span></h2><!--]--><!----> <p>通过 <code>image</code> 传入一张或多张参考图像。单张参考图像会保持完整:仅在超过 400 万像素时进行下采样,填充至 16 像素网格,且生成默认采用其尺寸,因此编辑结果与输入尺寸一致。多张参考图像则各自被调整至 100 万像素以内并裁剪至网格,生成默认采用第一张的尺寸。指定的 <code>height</code>/<code>width</code> 会覆盖默认值。请将 <em>目标</em> 图像描述为包含编辑内容的参考图像完整说明,而非单行指令;或者让 <a href="#prompt-upsampling">prompt upsampler</a> 根据指令和参考图像来撰写该说明:</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 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> 已复制</div><!----></button><!----></div> <pre class="language-python "><!---->night = pipe( | |
| prompt=<span class="hljs-string">"the same scene at night under a full moon"</span>, | |
| image=image, | |
| generator=torch.Generator().manual_seed(<span class="hljs-number">0</span>), | |
| output=[<span class="hljs-string">"images"</span>, <span class="hljs-string">"prompt"</span>], | |
| )<!----></pre></div><!----> <p>当提供多个参考图像时,请按照传入 <code>image</code> 的顺序,将它们称为第一张图像、第二张图像……;在图像模型中,标题将它们命名为 <code><ref_image_0></code>、<code><ref_image_1></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 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> 已复制</div><!----></button><!----></div> <pre class="language-python "><!---->combined = pipe( | |
| prompt=<span class="hljs-string">"the fox from the first image sitting on the wooden bench from the second image"</span>, | |
| image=[image, bench], | |
| height=<span class="hljs-number">768</span>, | |
| width=<span class="hljs-number">1360</span>, | |
| output=<span class="hljs-string">"images"</span>, | |
| )[<span class="hljs-number">0</span>]<!----></pre></div><!----> <p>传入 <code>height</code> 和/或 <code>width</code> 以覆盖推导出的尺寸。</p> <!--[1--><h2 class="relative group"><a id="layout-control" 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="#layout-control"><span><svg 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><!---->布局控制<!----></span></h2><!--]--><!----> <p>若要显式放置内容,请传入 <code>boxes</code>:为每个元素指定一个 <code>id</code>、简短的 <code>desc</code>,以及一个 <code>bbox</code>,其格式为 <code>[y_min, x_min, y_max, x_max]</code>,取值范围为 0-1(原点位于左上角),从而使其与输出尺寸无关。上采样器会在这些框周围生成描述,并以 <code><id></code> 的形式引用每个元素,最后附上框的表格,其中您提供的框将原样保留:</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 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> 已复制</div><!----></button><!----></div> <pre class="language-python "><!---->boxes = [ | |
| {<span class="hljs-string">"id"</span>: <span class="hljs-string">"balloon_1"</span>, <span class="hljs-string">"desc"</span>: <span class="hljs-string">"a red and yellow striped hot-air balloon"</span>, <span class="hljs-string">"bbox"</span>: [<span class="hljs-number">0.08</span>, <span class="hljs-number">0.06</span>, <span class="hljs-number">0.48</span>, <span class="hljs-number">0.38</span>]}, | |
| {<span class="hljs-string">"id"</span>: <span class="hljs-string">"balloon_2"</span>, <span class="hljs-string">"desc"</span>: <span class="hljs-string">"a small blue hot-air balloon"</span>, <span class="hljs-string">"bbox"</span>: [<span class="hljs-number">0.20</span>, <span class="hljs-number">0.62</span>, <span class="hljs-number">0.42</span>, <span class="hljs-number">0.80</span>]}, | |
| ] | |
| out = pipe(prompt=<span class="hljs-string">"hot-air balloons over a green valley at sunrise"</span>, boxes=boxes, output=[<span class="hljs-string">"images"</span>, <span class="hljs-string">"prompt"</span>])<!----></pre></div><!----> <p>在编辑时,一个框通过 <code>from</code> 指定其来源:<code>"target"</code> 配合 <code>tgt_bbox</code> 会在该位置添加一个新元素;<code>"ref_image_0"</code> 配合 <code>src_bbox</code> 和 <code>tgt_bbox</code> 会移动该参考图像中的一个元素(若两个框相同,则元素保持原位):</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 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> 已复制</div><!----></button><!----></div> <pre class="language-python "><!---->boxes = [{<span class="hljs-string">"id"</span>: <span class="hljs-string">"cactus"</span>, <span class="hljs-string">"desc"</span>: <span class="hljs-string">"a small potted cactus"</span>, <span class="hljs-string">"from"</span>: <span class="hljs-string">"target"</span>, <span class="hljs-string">"src_bbox"</span>: <span class="hljs-literal">None</span>, <span class="hljs-string">"tgt_bbox"</span>: [<span class="hljs-number">0.55</span>, <span class="hljs-number">0.62</span>, <span class="hljs-number">0.85</span>, <span class="hljs-number">0.85</span>]}] | |
| out = pipe(prompt=<span class="hljs-string">"add a small potted cactus in the marked spot"</span>, image=mug, boxes=boxes, output=[<span class="hljs-string">"images"</span>, <span class="hljs-string">"prompt"</span>]) | |
| boxes = [{<span class="hljs-string">"id"</span>: <span class="hljs-string">"knight"</span>, <span class="hljs-string">"desc"</span>: <span class="hljs-string">"the knight figure"</span>, <span class="hljs-string">"from"</span>: <span class="hljs-string">"ref_image_0"</span>, <span class="hljs-string">"src_bbox"</span>: [<span class="hljs-number">0.50</span>, <span class="hljs-number">0.15</span>, <span class="hljs-number">0.85</span>, <span class="hljs-number">0.35</span>], <span class="hljs-string">"tgt_bbox"</span>: [<span class="hljs-number">0.19</span>, <span class="hljs-number">0.06</span>, <span class="hljs-number">0.54</span>, <span class="hljs-number">0.26</span>]}] | |
| out = pipe(prompt=<span class="hljs-string">"move the knight up and to the left, keep the rest unchanged"</span>, image=knight, boxes=boxes, output=[<span class="hljs-string">"images"</span>, <span class="hljs-string">"prompt"</span>])<!----></pre></div><!----> <p>图像模型读取的说明文本中,元素以 <code><id></code> 标记,表格以 JSON 形式附加,且框的位置基于 0-1000 的画布;打印 <code>prompt</code> 的输出即可查看。若手动编写该形式的说明文本,则无需经过此步骤即可传入流水线,如 <a href="#prompt-upsampling">Prompt upsampling</a> 中所述。</p> <!--[1--><h2 class="relative group"><a id="guidance-scale" 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="#guidance-scale"><span><svg 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><!---->引导比例<!----></span></h2><!--]--><!----> <p>该 pipeline 通过其 <code>guider</code> 组件应用 classifier-free guidance,并且在调用时不接受 <code>guidance_scale</code> 参数。默认的 guider 是 <a href="/docs/diffusers/v0.43.0/zh/api/modular_diffusers/guiders#diffusers.ClassifierFreeGuidance">ClassifierFreeGuidance</a>,其 <code>guidance_scale=3.5</code>,这是两种工作流均推荐的设置。若要更改它,请创建一个更新后的 guider 副本:</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 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> 已复制</div><!----></button><!----></div> <pre class="language-python "><!---->pipe.guider = pipe.guider.new(guidance_scale=<span class="hljs-number">5.0</span>)<!----></pre></div><!----> <p>请参阅<a href="../../using-diffusers/guiders">guiders 指南</a>以切换到不同的引导方法。</p> <!--[1--><h2 class="relative group"><a id="step-distilled-model" 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="#step-distilled-model"><span><svg 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><!---->步数蒸馏模型<!----></span></h2><!--]--><!----> <p>步数蒸馏检查点可在 16 步内生成,且无需分类器自由引导。其 Transformer 是一个流映射:在每一步中,它都会被告知当前时间步以及该步所跳转到的目标时间步。它的加载方式与基础模型相同(该仓库的 <code>modular_model_index.json</code> 解析为 <a href="/docs/diffusers/v0.43.0/zh/api/pipelines/flux3#diffusers.Flux3DistilledModularPipeline">Flux3DistilledModularPipeline</a> 和 <a href="/docs/diffusers/v0.43.0/zh/api/pipelines/flux3#diffusers.Flux3DistilledAutoBlocks">Flux3DistilledAutoBlocks</a>);该流水线没有 <code>guider</code>,也不接受 <code>negative_prompt</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 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> 已复制</div><!----></button><!----></div> <pre class="language-python "><!---->pipe = ModularPipeline.from_pretrained(<span class="hljs-string">"black-forest-labs/flux-3-image-klein"</span>) | |
| pipe.load_components(dtype=torch.bfloat16) | |
| pipe.to(<span class="hljs-string">"cuda"</span>) | |
| image = pipe(prompt=prompt, generator=torch.Generator().manual_seed(<span class="hljs-number">0</span>), output=<span class="hljs-string">"images"</span>)[<span class="hljs-number">0</span>]<!----></pre></div><!----> <p>它支持与基础模型相同的两种工作流。<code>num_inference_steps</code> 的默认值为 16;该模型经过 4 到 16 步的训练。</p> <!--[1--><h2 class="relative group"><a id="reproducible-generation" 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="#reproducible-generation"><span><svg 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><!---->可复现的生成<!----></span></h2><!--]--><!----> <p><code>generator</code> 用于固定流水线所采样的噪声(参见 <a href="../../using-diffusers/reusing_seeds">可复现性</a>)。若要复用其他位置采样的噪声,请将其作为 <code>latents</code> 传入,其形状应为 <code>(batch_size, 128, height // 16, width // 16)</code>。例如,<a href="https://github.com/black-forest-labs/flux-3" rel="nofollow">BFL 实现</a> 在 CPU 上将噪声采样为图像 token,并将其转换为 bf16;对于给定的种子,这可以复现其采样结果:</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 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> 已复制</div><!----></button><!----></div> <pre class="language-python "><!---->height, width = <span class="hljs-number">1024</span>, <span class="hljs-number">1024</span> | |
| h, w = height // <span class="hljs-number">16</span>, width // <span class="hljs-number">16</span> | |
| noise = torch.randn((h * w, <span class="hljs-number">128</span>), generator=torch.Generator().manual_seed(<span class="hljs-number">0</span>)).to(torch.bfloat16) | |
| latents = noise.T.reshape(<span class="hljs-number">1</span>, <span class="hljs-number">128</span>, h, w) | |
| image = pipe(prompt=prompt, latents=latents, height=height, width=width, output=<span class="hljs-string">"images"</span>)[<span class="hljs-number">0</span>]<!----></pre></div><!----> <!--[1--><h2 class="relative group"><a id="workflows" 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="#workflows"><span><svg 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><!---->工作流程<!----></span></h2><!--]--><!----> <p>该流水线支持两种工作流:<code>text2image</code> 和 <code>image_conditioned</code>;实际执行哪一种由输入决定(传入 <code>image</code> 即选择图生图)。<a href="/docs/diffusers/v0.43.0/zh/api/modular_diffusers/pipeline#diffusers.ModularPipeline.from_pretrained">from_pretrained()</a> 也可以通过 <code>workflow</code> 参数为单一工作流构建流水线,此时它只会加载该工作流所需的组件:</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 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> 已复制</div><!----></button><!----></div> <pre class="language-python "><!----><span class="hljs-keyword">from</span> diffusers.modular_pipelines <span class="hljs-keyword">import</span> Flux3AutoBlocks | |
| <span class="hljs-built_in">print</span>(Flux3AutoBlocks().available_workflows) | |
| <span class="hljs-comment"># ["text2image", "image_conditioned"]</span> | |
| t2i = ModularPipeline.from_pretrained(<span class="hljs-string">"black-forest-labs/flux-3-image-klein-base"</span>, workflow=<span class="hljs-string">"text2image"</span>) | |
| <span class="hljs-built_in">print</span>(t2i.blocks.doc)<!----></pre></div><!----> <!--[1--><h2 class="relative group"><a id="per-step-callbacks" 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="#per-step-callbacks"><span><svg 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><!---->逐步回调<!----></span></h2><!--]--><!----> <p>若要使用现有的回调函数,请在去噪循环中插入一个代码块。<a href="../../using-diffusers/callback#modular-pipelines">callbacks 指南</a>提供了一个使用 <code>CallbackOnStepEnd</code> 代码块和 <code>walk_denoise_loops</code> 辅助函数的简单示例:</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 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> 已复制</div><!----></button><!----></div> <pre class="language-python "><!---->blocks = Flux3AutoBlocks() <span class="hljs-comment"># or Flux3DistilledAutoBlocks</span> | |
| <span class="hljs-keyword">for</span> loop <span class="hljs-keyword">in</span> walk_denoise_loops(blocks): | |
| loop.sub_blocks[<span class="hljs-string">"callback"</span>] = CallbackOnStepEnd() | |
| pipe = blocks.init_pipeline(<span class="hljs-string">"black-forest-labs/flux-3-image-klein-base"</span>) | |
| image = pipe(prompt=caption, callback_on_step_end=my_callback, output=<span class="hljs-string">"images"</span>)[<span class="hljs-number">0</span>]<!----></pre></div><!----> <p>若要提前停止,请从回调中抛出异常;不存在 <code>interrupt</code> 标志。</p> <!--[1--><h2 class="relative group"><a id="memory" 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="#memory"><span><svg 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><!---->内存<!----></span></h2><!--]--><!----> <p>以 bf16 格式加载所有内容大约需要 39 GB 的权重:transformer 为 5.5B 参数,Qwen3-VL-4B 文本编码器为 4B,<a href="#prompt-upsampling">prompt upsampler</a> 为 9B(单独占用约 19 GB),图像 VAE 则以 fp32 格式加载。如果构建的 pipeline 不包含 upsampler 步骤(参见此处),则加载量约为 20 GB。对于显存较小的 GPU,请启用组件卸载(component offloading);管理器会在每个步骤需要时,在 CPU 和 GPU 之间移动组件,upsampler 也包括在内:</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 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> 已复制</div><!----></button><!----></div> <pre class="language-python "><!----><span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> ComponentsManager, ModularPipeline | |
| manager = ComponentsManager() | |
| pipe = ModularPipeline.from_pretrained(<span class="hljs-string">"black-forest-labs/flux-3-image-klein-base"</span>, components_manager=manager) | |
| pipe.load_components(dtype={<span class="hljs-string">"vae"</span>: torch.float32, <span class="hljs-string">"default"</span>: torch.bfloat16}) | |
| manager.enable_auto_cpu_offload(device=<span class="hljs-string">"cuda"</span>)<!----></pre></div><!----> <!--[1--><h2 class="relative group"><a id="single-file-loading" 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="#single-file-loading"><span><svg 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><!---->单文件加载<!----></span></h2><!--]--><!----> <p><a href="/docs/diffusers/v0.43.0/zh/api/models/flux3_transformer#diffusers.Flux3Transformer2DModel">Flux3Transformer2DModel</a> 通过 <a href="/docs/diffusers/v0.43.0/zh/api/loaders/single_file#diffusers.loaders.FromOriginalModelMixin.from_single_file">from_single_file()</a> 直接从 Black Forest Labs 的模型包中加载 transformer。该包中的 <code>config.json</code> 不会被读取;请通过 <code>config</code> 和 <code>subfolder</code> 传入 diffusers 配置。</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 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> 已复制</div><!----></button><!----></div> <pre class="language-python "><!----><span class="hljs-keyword">import</span> torch | |
| <span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> Flux3Transformer2DModel, ModularPipeline | |
| transformer = Flux3Transformer2DModel.from_single_file( | |
| <span class="hljs-string">"https://huggingface.co/black-forest-labs/flux-3-image-klein-base/blob/main/model.safetensors"</span>, | |
| config=<span class="hljs-string">"black-forest-labs/flux-3-image-klein-base"</span>, | |
| subfolder=<span class="hljs-string">"transformer"</span>, | |
| dtype=torch.bfloat16, | |
| ) | |
| pipe = ModularPipeline.from_pretrained(<span class="hljs-string">"black-forest-labs/flux-3-image-klein-base"</span>) | |
| pipe.load_components(dtype={<span class="hljs-string">"vae"</span>: torch.float32, <span class="hljs-string">"default"</span>: torch.bfloat16}) | |
| pipe.update_components(transformer=transformer) | |
| pipe.to(<span class="hljs-string">"cuda"</span>)<!----></pre></div><!----> <p>首先按名称加载其他组件:如果不指定 <code>names</code>,<a href="/docs/diffusers/v0.43.0/zh/api/modular_diffusers/pipeline#diffusers.ModularPipeline.load_components">load_components()</a> 也会从流水线仓库中加载 transformer。</p> <p>同样支持加载 FP8r 和 NVFP4 单文件检查点。两者均通过相同的调用方式加载,并使用 <a href="/docs/diffusers/v0.43.0/zh/api/quantization#diffusers.BFLQuantizationConfig">BFLQuantizationConfig</a>。有关每种方案的 kernel 要求,请参阅 <a href="../../quantization/bfl">BFL 量化指南</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 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> 已复制</div><!----></button><!----></div> <pre class="language-python "><!----><span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> BFLQuantizationConfig | |
| transformer = Flux3Transformer2DModel.from_single_file( | |
| <span class="hljs-string">"https://huggingface.co/black-forest-labs/flux-3-image-klein-base/blob/main/variants/nvfp4/model.safetensors"</span>, | |
| config=<span class="hljs-string">"black-forest-labs/flux-3-image-klein-base"</span>, | |
| subfolder=<span class="hljs-string">"transformer"</span>, | |
| quantization_config=BFLQuantizationConfig(compute_dtype=torch.bfloat16), | |
| dtype=torch.bfloat16, | |
| )<!----></pre></div><!----> <p>加载步数蒸馏检查点时,我们需要它们的配置(<code>flow_map=True</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 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> 已复制</div><!----></button><!----></div> <pre class="language-python "><!---->transformer = Flux3Transformer2DModel.from_single_file( | |
| <span class="hljs-string">"https://huggingface.co/black-forest-labs/flux-3-image-klein/blob/main/variants/nvfp4/model.safetensors"</span>, | |
| config=<span class="hljs-string">"black-forest-labs/flux-3-image-klein"</span>, | |
| subfolder=<span class="hljs-string">"transformer"</span>, | |
| quantization_config=BFLQuantizationConfig(compute_dtype=torch.bfloat16), | |
| dtype=torch.bfloat16, | |
| )<!----></pre></div><!----> <!--[1--><h2 class="relative group"><a id="diffusers.Flux3ModularPipeline" 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="#diffusers.Flux3ModularPipeline"><span><svg 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><!---->Flux3ModularPipeline<!----></span></h2><!--]--><!----> <div class="docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"><div><span class="group flex space-x-1.5 items-center text-gray-800 bg-gradient-to-r rounded-tr-lg -mt-4 -ml-4 pt-3 px-2.5" id="diffusers.Flux3ModularPipeline"><!----><h3 class="!m-0"><span class="flex-1 break-all md:text-lg bg-gradient-to-r px-2.5 py-1.5 rounded-xl from-indigo-50/70 to-white dark:from-gray-900 dark:to-gray-950 dark:text-indigo-300 text-indigo-700"><svg class="mr-1.5 text-indigo-500 inline-block -mt-0.5" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" focusable="false" role="img" width=".8em" height=".8em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 24 24"><path class="uim-quaternary" d="M20.23 7.24L12 12L3.77 7.24a1.98 1.98 0 0 1 .7-.71L11 2.76c.62-.35 1.38-.35 2 0l6.53 3.77c.29.173.531.418.7.71z" opacity=".25" fill="currentColor"></path><path class="uim-tertiary" d="M12 12v9.5a2.09 2.09 0 0 1-.91-.21L4.5 17.48a2.003 2.003 0 0 1-1-1.73v-7.5a2.06 2.06 0 0 1 .27-1.01L12 12z" opacity=".5" fill="currentColor"></path><path class="uim-primary" d="M20.5 8.25v7.5a2.003 2.003 0 0 1-1 1.73l-6.62 3.82c-.275.13-.576.198-.88.2V12l8.23-4.76c.175.308.268.656.27 1.01z" fill="currentColor"></path></svg><span class="font-light">class</span> <span class="font-medium">diffusers.</span><span class="font-semibold">Flux3ModularPipeline</span></span></h3><!----> <a id="diffusers.Flux3ModularPipeline" class="header-link invisible with-hover:group-hover:visible pr-2" href="#diffusers.Flux3ModularPipeline"><svg 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><!----></a> <!--[0--><a class="!ml-auto !text-gray-400 !no-underline text-sm flex items-center" href="https://github.com/huggingface/diffusers/blob/v0.43.0/src/diffusers/modular_pipelines/flux3/modular_pipeline.py#L24" target="_blank"><span><</span> <span class="hidden md:block mx-0.5 hover:!underline">source</span> <span>></span></a><!--]--></span> <!--[0--><p class="font-mono text-xs md:text-sm !leading-relaxed !my-6"><span>(</span> <!--[--><span class="comma cursor-default"><span class="rounded hover:bg-black hover:text-white dark:hover:bg-white dark:hover:text-black">blocks<span class="opacity-60">: diffusers.modular_pipelines.modular_pipeline.ModularPipelineBlocks | None = None</span></span></span><span class="comma cursor-default"><span class="rounded hover:bg-black hover:text-white dark:hover:bg-white dark:hover:text-black">pretrained_model_name_or_path<span class="opacity-60">: str | os.PathLike | None = None</span></span></span><span class="comma cursor-default"><span class="rounded hover:bg-black hover:text-white dark:hover:bg-white dark:hover:text-black">components_manager<span class="opacity-60">: diffusers.modular_pipelines.components_manager.ComponentsManager | None = None</span></span></span><span class="comma cursor-default"><span class="rounded hover:bg-black hover:text-white dark:hover:bg-white dark:hover:text-black">collection<span class="opacity-60">: str | None = None</span></span></span><span class="comma cursor-default"><span class="rounded hover:bg-black hover:text-white dark:hover:bg-white dark:hover:text-black">workflow<span class="opacity-60">: str | None = None</span></span></span><span class="comma cursor-default"><span class="rounded hover:bg-black hover:text-white dark:hover:bg-white dark:hover:text-black">modular_config_dict<span class="opacity-60">: dict[str, typing.Any] | None = None</span></span></span><span class="comma cursor-default"><span class="rounded hover:bg-black hover:text-white dark:hover:bg-white dark:hover:text-black">config_dict<span class="opacity-60">: dict[str, typing.Any] | None = None</span></span></span><span class="comma cursor-default"><span class="rounded hover:bg-black hover:text-white dark:hover:bg-white dark:hover:text-black">**kwargs<span class="opacity-60"></span></span></span><!--]--> <span>)</span> <!--[-1--><!--]--></p><!--]--> <div class="!mb-10 relative docstring-details "><!--[-1--><!--]--> <!--[-1--><!--]--> <!--[-1--><!--]--> <!--[-1--><!--]--> <!--[-1--><!--]--></div></div><!----> <p>用于 FLUX.3 的 ModularPipeline。</p></div> <!--[1--><h2 class="relative group"><a id="diffusers.Flux3AutoBlocks" 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="#diffusers.Flux3AutoBlocks"><span><svg 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><!---->Flux3AutoBlocks<!----></span></h2><!--]--><!----> <div class="docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"><div><span class="group flex space-x-1.5 items-center text-gray-800 bg-gradient-to-r rounded-tr-lg -mt-4 -ml-4 pt-3 px-2.5" id="diffusers.Flux3AutoBlocks"><!----><h3 class="!m-0"><span class="flex-1 break-all md:text-lg bg-gradient-to-r px-2.5 py-1.5 rounded-xl from-indigo-50/70 to-white dark:from-gray-900 dark:to-gray-950 dark:text-indigo-300 text-indigo-700"><svg class="mr-1.5 text-indigo-500 inline-block -mt-0.5" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" focusable="false" role="img" width=".8em" height=".8em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 24 24"><path class="uim-quaternary" d="M20.23 7.24L12 12L3.77 7.24a1.98 1.98 0 0 1 .7-.71L11 2.76c.62-.35 1.38-.35 2 0l6.53 3.77c.29.173.531.418.7.71z" opacity=".25" fill="currentColor"></path><path class="uim-tertiary" d="M12 12v9.5a2.09 2.09 0 0 1-.91-.21L4.5 17.48a2.003 2.003 0 0 1-1-1.73v-7.5a2.06 2.06 0 0 1 .27-1.01L12 12z" opacity=".5" fill="currentColor"></path><path class="uim-primary" d="M20.5 8.25v7.5a2.003 2.003 0 0 1-1 1.73l-6.62 3.82c-.275.13-.576.198-.88.2V12l8.23-4.76c.175.308.268.656.27 1.01z" fill="currentColor"></path></svg><span class="font-light">class</span> <span class="font-medium">diffusers.</span><span class="font-semibold">Flux3AutoBlocks</span></span></h3><!----> <a id="diffusers.Flux3AutoBlocks" class="header-link invisible with-hover:group-hover:visible pr-2" href="#diffusers.Flux3AutoBlocks"><svg 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><!----></a> <!--[0--><a class="!ml-auto !text-gray-400 !no-underline text-sm flex items-center" href="https://github.com/huggingface/diffusers/blob/v0.43.0/src/diffusers/modular_pipelines/flux3/modular_blocks_flux3.py#L339" target="_blank"><span><</span> <span class="hidden md:block mx-0.5 hover:!underline">source</span> <span>></span></a><!--]--></span> <!--[0--><p class="font-mono text-xs md:text-sm !leading-relaxed !my-6"><span>(</span> <!--[--><!--]--> <span>)</span> <!--[-1--><!--]--></p><!--]--> <div class="!mb-10 relative docstring-details "><!--[-1--><!--]--> <!--[-1--><!--]--> <!--[-1--><!--]--> <!--[-1--><!--]--> <!--[-1--><!--]--></div></div><!----> <p>使用 FLUX.3 执行文本生成图像和图像生成图像任务的 Auto 块。<code>prompt</code> 会先经过提示词上采样器;若弹出 <code>prompt_upsampler</code> 步骤,则会按原样对其进行编码。</p> <p>支持的工作流:</p> <ul><li><code>text2image</code>:需要 <code>prompt</code></li> <li><code>image_conditioned</code>:需要 <code>prompt</code>、<code>image</code></li></ul> <p>组件: | |
| prompt_upsampler (<code>Qwen3_5ForConditionalGeneration</code>):提示词增强器。prompt_upsampler_processor | |
| (<code>AutoProcessor</code>):其对应的处理器。text_encoder (<code>Qwen3VLForConditionalGeneration</code>) tokenizer | |
| (<code>AutoProcessor</code>) image_processor (<code>Flux3ImageProcessor</code>) vae (<code>AutoencoderKLFlux2</code>) transformer | |
| (<code>Flux3Transformer2DModel</code>) scheduler (<code>FlowMatchEulerDiscreteScheduler</code>) guider (<code>ClassifierFreeGuidance</code>)</p> <p>输入: | |
| prompt (<code>str</code>): | |
| 用于扩展为完整描述的编辑指令(每次调用一个)。 | |
| image (<code>Image | list</code>,<em>可选</em>): | |
| 编辑的参考图像,最多 10 张;与流水线接收的图像相同。 | |
| boxes (<code>list</code>,<em>可选</em>): | |
| 布局框,每个框为 {‘id’, ‘desc’, ‘from’, ‘src_bbox’, ‘tgt_bbox’},其中框的坐标为 [y_min, x_min, y_max, x_max],取值范围为 0-1:‘from’: ‘target’ 配合 <code>tgt_bbox</code> 表示在该位置添加一个元素;‘from’: ‘ref_image_0’,… 配合 <code>src_bbox</code> 和 <code>tgt_bbox</code> 表示从参考图像中移动一个元素。 | |
| max_new_tokens (<code>int</code>,<em>可选</em>,默认为 3584): | |
| upsampler 为描述生成的最大 token 数。 | |
| negative_prompt (<code>str</code>,<em>可选</em>): | |
| 用于不引导生成的提示词。 | |
| max_sequence_length (<code>int</code>,<em>可选</em>,默认为 8192): | |
| 提示词 token 的最大数量;较长的提示词将被截断。 | |
| text_encoder_out_layers (<code>tuple</code>,<em>可选</em>,默认为 (4, 8, 12, 16, 20, 24, 28, 32)): | |
| 文本编码器隐藏层的索引,这些层的状态将被堆叠以形成文本嵌入。 | |
| num_images_per_prompt (<code>int</code>,<em>可选</em>,默认为 1): | |
| 每个提示词生成的图像数量。 | |
| image_cond_latents (<code>list</code>,<em>可选</em>): | |
| 参考图像的归一化打包潜变量,每张图像对应一个形状为 (B, C, height_i//16, width_i//16) 的张量。 | |
| height (<code>int</code>,<em>可选</em>,默认为 None 或 1024,具体取决于工作流): | |
| 生成图像的高度(像素);默认为第一张参考图像自身的高度。 | |
| width (<code>int</code>,<em>可选</em>,默认为 None 或 1024,具体取决于工作流): | |
| 生成图像的宽度(像素);默认为第一张参考图像自身的宽度。 | |
| reference_crop (<code>list | NoneType</code>,<em>可选</em>): | |
| 对于每个参考图像,其在填充画布内的框,或 None,来自预处理步骤。 | |
| latents (<code>Tensor</code>): | |
| 用于图像生成的预生成噪声潜变量。 | |
| generator (<code>Generator</code>,<em>可选</em>): | |
| 用于确定性生成的 Torch 生成器。 | |
| num_inference_steps (<code>int</code>): | |
| 去噪步骤数。 | |
| timesteps (<code>Tensor</code>): | |
| 去噪过程的时间步。(提供时会覆盖 <code>num_inference_steps</code>) | |
| sigmas (<code>list</code>,<em>可选</em>): | |
| 去噪过程的自定义 sigmas。(提供时会覆盖 <code>num_inference_steps</code>) | |
| *<em>denoiser_input_fields (<code>None</code>,</em>optional<em>): | |
| 去噪器的条件模型输入:例如 prompt_embeds、negative_prompt_embeds 等。 | |
| attention_kwargs (<code>dict</code>,</em>optional<em>): | |
| 注意力处理器的额外 kwargs。 | |
| output_type (<code>str</code>,</em>optional*,默认为 pil): | |
| 输出格式:‘pil’、‘np’、‘pt’。</p> <p>输出: | |
| images (<code>list</code>): | |
| 生成的图像。 | |
| latents (<code>Tensor</code>): | |
| 去噪后的潜在变量。</p></div> <!--[1--><h2 class="relative group"><a id="diffusers.Flux3DistilledAutoBlocks" 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="#diffusers.Flux3DistilledAutoBlocks"><span><svg 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><!---->Flux3DistilledAutoBlocks<!----></span></h2><!--]--><!----> <div class="docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"><div><span class="group flex space-x-1.5 items-center text-gray-800 bg-gradient-to-r rounded-tr-lg -mt-4 -ml-4 pt-3 px-2.5" id="diffusers.Flux3DistilledAutoBlocks"><!----><h3 class="!m-0"><span class="flex-1 break-all md:text-lg bg-gradient-to-r px-2.5 py-1.5 rounded-xl from-indigo-50/70 to-white dark:from-gray-900 dark:to-gray-950 dark:text-indigo-300 text-indigo-700"><svg class="mr-1.5 text-indigo-500 inline-block -mt-0.5" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" focusable="false" role="img" width=".8em" height=".8em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 24 24"><path class="uim-quaternary" d="M20.23 7.24L12 12L3.77 7.24a1.98 1.98 0 0 1 .7-.71L11 2.76c.62-.35 1.38-.35 2 0l6.53 3.77c.29.173.531.418.7.71z" opacity=".25" fill="currentColor"></path><path class="uim-tertiary" d="M12 12v9.5a2.09 2.09 0 0 1-.91-.21L4.5 17.48a2.003 2.003 0 0 1-1-1.73v-7.5a2.06 2.06 0 0 1 .27-1.01L12 12z" opacity=".5" fill="currentColor"></path><path class="uim-primary" d="M20.5 8.25v7.5a2.003 2.003 0 0 1-1 1.73l-6.62 3.82c-.275.13-.576.198-.88.2V12l8.23-4.76c.175.308.268.656.27 1.01z" fill="currentColor"></path></svg><span class="font-light">class</span> <span class="font-medium">diffusers.</span><span class="font-semibold">Flux3DistilledAutoBlocks</span></span></h3><!----> <a id="diffusers.Flux3DistilledAutoBlocks" class="header-link invisible with-hover:group-hover:visible pr-2" href="#diffusers.Flux3DistilledAutoBlocks"><svg 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><!----></a> <!--[0--><a class="!ml-auto !text-gray-400 !no-underline text-sm flex items-center" href="https://github.com/huggingface/diffusers/blob/v0.43.0/src/diffusers/modular_pipelines/flux3/modular_blocks_flux3_distilled.py#L258" target="_blank"><span><</span> <span class="hidden md:block mx-0.5 hover:!underline">source</span> <span>></span></a><!--]--></span> <!--[0--><p class="font-mono text-xs md:text-sm !leading-relaxed !my-6"><span>(</span> <!--[--><!--]--> <span>)</span> <!--[-1--><!--]--></p><!--]--> <div class="!mb-10 relative docstring-details "><!--[-1--><!--]--> <!--[-1--><!--]--> <!--[-1--><!--]--> <!--[-1--><!--]--> <!--[-1--><!--]--></div></div><!----> <p>Auto 块使用步数蒸馏的 FLUX.3 模型(16 步,无分类器自由引导)执行文本到图像和图像到图像生成。<code>prompt</code> 会先经过提示词增强器;若弹出 <code>prompt_upsampler</code> 步骤,则按原样对其进行编码。</p> <p>支持的工作流:</p> <ul><li><code>text2image</code>:需要 <code>prompt</code></li> <li><code>image_conditioned</code>:需要 <code>prompt</code>、<code>image</code></li></ul> <p>组件: | |
| prompt_upsampler (<code>Qwen3_5ForConditionalGeneration</code>):提示词增强器。prompt_upsampler_processor | |
| (<code>AutoProcessor</code>):其对应的处理器。text_encoder (<code>Qwen3VLForConditionalGeneration</code>) tokenizer | |
| (<code>AutoProcessor</code>) image_processor (<code>Flux3ImageProcessor</code>) vae (<code>AutoencoderKLFlux2</code>) transformer | |
| (<code>Flux3Transformer2DModel</code>) scheduler (<code>FlowMatchEulerDiscreteScheduler</code>)</p> <p>输入: | |
| prompt (<code>str</code>): | |
| 用于扩展为完整描述的编辑指令(每次调用一个)。 | |
| image (<code>Image | list</code>,<em>可选</em>): | |
| 编辑的参考图像,最多 10 张;与流水线接收的图像相同。 | |
| boxes (<code>list</code>,<em>可选</em>): | |
| 布局框,每个框为 {‘id’, ‘desc’, ‘from’, ‘src_bbox’, ‘tgt_bbox’},其中框的坐标为 [y_min, x_min, y_max, x_max],取值范围为 0-1:‘from’: ‘target’ 且带有 <code>tgt_bbox</code> 表示在该位置添加一个元素;‘from’: ‘ref_image_0’,… 且带有 <code>src_bbox</code> 和 <code>tgt_bbox</code> 表示从参考图像中移动一个元素。 | |
| max_new_tokens (<code>int</code>,<em>可选</em>,默认为 3584): | |
| 描述生成器为描述生成的最大 token 数。 | |
| max_sequence_length (<code>int</code>,<em>可选</em>,默认为 8192): | |
| 提示词 token 的最大数量;较长的提示词将被截断。 | |
| text_encoder_out_layers (<code>tuple</code>,<em>可选</em>,默认为 (4, 8, 12, 16, 20, 24, 28, 32)): | |
| 文本编码器隐藏层的索引,这些层的状态将被堆叠以形成文本嵌入 | |
| num_images_per_prompt (<code>int</code>,<em>可选</em>,默认为 1): | |
| 每个提示词生成的图像数量。 | |
| image_cond_latents (<code>list</code>,<em>可选</em>): | |
| 参考图像的归一化打包潜在变量,每张图像对应一个形状为 (B, C, height_i//16, width_i//16) 的张量 | |
| height (<code>int</code>,<em>可选</em>,默认为 None 或 1024,具体取决于工作流): | |
| 生成图像的高度(像素);默认为第一张参考图像自身的高度 | |
| width (<code>int</code>,<em>可选</em>,默认为 None 或 1024,具体取决于工作流): | |
| 生成图像的宽度(像素);默认为第一张参考图像自身的宽度 | |
| reference_crop (<code>list | NoneType</code>,<em>可选</em>): | |
| 对于每个参考图像,其在填充画布内的框,或 None,来自预处理步骤 | |
| latents (<code>Tensor</code>): | |
| 用于图像生成的预生成噪声潜在变量。 | |
| generator (<code>Generator</code>,<em>可选</em>): | |
| 用于确定性生成的 Torch 生成器。 | |
| num_inference_steps (<code>int</code>): | |
| 去噪步骤数。 | |
| timesteps (<code>Tensor</code>): | |
| 去噪过程的时间步。(提供时会覆盖 <code>num_inference_steps</code>) | |
| sigmas (<code>list</code>,<em>可选</em>): | |
| 去噪过程的自定义 sigmas。(提供时会覆盖 <code>num_inference_steps</code>) | |
| *<em>denoiser_input_fields (<code>None</code>,</em>optional<em>): | |
| 去噪器的条件模型输入:例如 prompt_embeds、negative_prompt_embeds 等。 | |
| output_type (<code>str</code>,</em>optional*,默认为 pil): | |
| 输出格式:‘pil’、‘np’、‘pt’。</p> <p>输出: | |
| images (<code>list</code>): | |
| 生成的图像。 | |
| latents (<code>Tensor</code>): | |
| 去噪后的潜在变量。</p></div> <!--[1--><h2 class="relative group"><a id="diffusers.Flux3DistilledModularPipeline" 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="#diffusers.Flux3DistilledModularPipeline"><span><svg 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><!---->Flux3DistilledModularPipeline<!----></span></h2><!--]--><!----> <div class="docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"><div><span class="group flex space-x-1.5 items-center text-gray-800 bg-gradient-to-r rounded-tr-lg -mt-4 -ml-4 pt-3 px-2.5" id="diffusers.Flux3DistilledModularPipeline"><!----><h3 class="!m-0"><span class="flex-1 break-all md:text-lg bg-gradient-to-r px-2.5 py-1.5 rounded-xl from-indigo-50/70 to-white dark:from-gray-900 dark:to-gray-950 dark:text-indigo-300 text-indigo-700"><svg class="mr-1.5 text-indigo-500 inline-block -mt-0.5" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" focusable="false" role="img" width=".8em" height=".8em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 24 24"><path class="uim-quaternary" d="M20.23 7.24L12 12L3.77 7.24a1.98 1.98 0 0 1 .7-.71L11 2.76c.62-.35 1.38-.35 2 0l6.53 3.77c.29.173.531.418.7.71z" opacity=".25" fill="currentColor"></path><path class="uim-tertiary" d="M12 12v9.5a2.09 2.09 0 0 1-.91-.21L4.5 17.48a2.003 2.003 0 0 1-1-1.73v-7.5a2.06 2.06 0 0 1 .27-1.01L12 12z" opacity=".5" fill="currentColor"></path><path class="uim-primary" d="M20.5 8.25v7.5a2.003 2.003 0 0 1-1 1.73l-6.62 3.82c-.275.13-.576.198-.88.2V12l8.23-4.76c.175.308.268.656.27 1.01z" fill="currentColor"></path></svg><span class="font-light">class</span> <span class="font-medium">diffusers.</span><span class="font-semibold">Flux3DistilledModularPipeline</span></span></h3><!----> <a id="diffusers.Flux3DistilledModularPipeline" class="header-link invisible with-hover:group-hover:visible pr-2" href="#diffusers.Flux3DistilledModularPipeline"><svg 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><!----></a> <!--[0--><a class="!ml-auto !text-gray-400 !no-underline text-sm flex items-center" href="https://github.com/huggingface/diffusers/blob/v0.43.0/src/diffusers/modular_pipelines/flux3/modular_pipeline.py#L52" target="_blank"><span><</span> <span class="hidden md:block mx-0.5 hover:!underline">source</span> <span>></span></a><!--]--></span> <!--[0--><p class="font-mono text-xs md:text-sm !leading-relaxed !my-6"><span>(</span> <!--[--><span class="comma cursor-default"><span class="rounded hover:bg-black hover:text-white dark:hover:bg-white dark:hover:text-black">blocks<span class="opacity-60">: diffusers.modular_pipelines.modular_pipeline.ModularPipelineBlocks | None = None</span></span></span><span class="comma cursor-default"><span class="rounded hover:bg-black hover:text-white dark:hover:bg-white dark:hover:text-black">pretrained_model_name_or_path<span class="opacity-60">: str | os.PathLike | None = None</span></span></span><span class="comma cursor-default"><span class="rounded hover:bg-black hover:text-white dark:hover:bg-white dark:hover:text-black">components_manager<span class="opacity-60">: diffusers.modular_pipelines.components_manager.ComponentsManager | None = None</span></span></span><span class="comma cursor-default"><span class="rounded hover:bg-black hover:text-white dark:hover:bg-white dark:hover:text-black">collection<span class="opacity-60">: str | None = None</span></span></span><span class="comma cursor-default"><span class="rounded hover:bg-black hover:text-white dark:hover:bg-white dark:hover:text-black">workflow<span class="opacity-60">: str | None = None</span></span></span><span class="comma cursor-default"><span class="rounded hover:bg-black hover:text-white dark:hover:bg-white dark:hover:text-black">modular_config_dict<span class="opacity-60">: dict[str, typing.Any] | None = None</span></span></span><span class="comma cursor-default"><span class="rounded hover:bg-black hover:text-white dark:hover:bg-white dark:hover:text-black">config_dict<span class="opacity-60">: dict[str, typing.Any] | None = None</span></span></span><span class="comma cursor-default"><span class="rounded hover:bg-black hover:text-white dark:hover:bg-white dark:hover:text-black">**kwargs<span class="opacity-60"></span></span></span><!--]--> <span>)</span> <!--[-1--><!--]--></p><!--]--> <div class="!mb-10 relative docstring-details "><!--[-1--><!--]--> <!--[-1--><!--]--> <!--[-1--><!--]--> <!--[-1--><!--]--> <!--[-1--><!--]--></div></div><!----> <p>用于步数蒸馏 FLUX.3 模型(16 步,无分类器自由引导)的 ModularPipeline。</p></div> <a class="!text-gray-400 !no-underline text-sm flex items-center not-prose mt-4" href="https://github.com/huggingface/diffusers/blob/main/docs/source/en/api/pipelines/flux3.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><span class="underline">Update</span> on GitHub</span></a><!----> <p></p><!--]--><!----><!--]--><!--]--><!--]--> <!--[-1--><!--]--><!--]--> | |
| <script> | |
| { | |
| __sveltekit_ywq9g6 = { | |
| base: "/docs/diffusers/v0.43.0/en", | |
| assets: "/docs/diffusers/v0.43.0/en" | |
| }; | |
| const element = document.currentScript.parentElement; | |
| Promise.all([ | |
| import("/docs/diffusers/v0.43.0/en/_app/immutable/entry/start.1gaXe03k.js"), | |
| import("/docs/diffusers/v0.43.0/en/_app/immutable/entry/app.BwdYRbCn.js") | |
| ]).then(([kit, app]) => { | |
| kit.start(app, element, { | |
| node_ids: [0, 180], | |
| data: [null,null], | |
| form: null, | |
| error: null | |
| }); | |
| }); | |
| } | |
| </script> | |
Xet Storage Details
- Size:
- 95.9 kB
- Xet hash:
- a7bbdbe3cd926f6490e1c7f55c176fdf4a2ab427ed29647ef0fb5e99290e7c82
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.