diffusers/pr_13881/zh/_app/immutable/nodes/20.D-vCVVM1.js

import"../chunks/DsnmJJEf.js";import{i as u,h as w,C as B,H as I,b as k,a,E as G,s as _}from"../chunks/DdZvggmf.js";import{p as R,o as Q,s as l,f as J,a as e,b as C,c as t,n as E}from"../chunks/BbekZcyp.js";import{H as U}from"../chunks/BcnRgdDK.js";const Z='{"title":"顺序管道块","local":"顺序管道块","sections":[],"depth":1}';var Y=t('<meta name="hf:doc:metadata"/>'),X=t("<!> <!>",1),f=t("<p></p> <!> <!> <p><code>SequentialPipelineBlocks</code> 是一种多块类型，它将其他 <code>ModularPipelineBlocks</code> 按顺序组合在一起。数据通过 <code>intermediate_inputs</code> 和 <code>intermediate_outputs</code> 线性地从一个块流向下一个块。<code>SequentialPipelineBlocks</code> 中的每个块通常代表管道中的一个步骤，通过组合它们，您逐步构建一个管道。</p> <p>本指南向您展示如何将两个块连接成一个 <code>SequentialPipelineBlocks</code>。</p> <p>创建两个 <code>ModularPipelineBlocks</code>。第一个块 <code>InputBlock</code> 输出一个 <code>batch_size</code> 值，第二个块 <code>ImageEncoderBlock</code> 使用 <code>batch_size</code> 作为 <code>intermediate_inputs</code>。</p> <!> <p>通过定义一个<code>InsertableDict</code>来连接两个块，将块名称映射到块实例。块按照它们在<code>blocks_dict</code>中注册的顺序执行。</p> <p>使用<code>from_blocks_dict()</code>来创建一个<code>SequentialPipelineBlocks</code>。</p> <!> <p>通过调用<code>blocks</code>来检查<code>SequentialPipelineBlocks</code>中的子块，要获取更多关于输入和输出的详细信息，可以访问<code>docs</code>属性。</p> <!> <!> <p></p>",1);function F(T,b){R(b,!1),Q(()=>{new URLSearchParams(window.location.search).get("fw")}),u();var p=f();w("bixetp",s=>{var M=Y();_(M,"content",Z),e(s,M)});var c=l(J(p),2);B(c,{containerStyle:"float: right; margin-left: 10px; display: inline-flex; position: relative; z-index: 10;"});var j=l(c,2);I(j,{title:"顺序管道块",local:"顺序管道块",headingTag:"h1"});var o=l(j,8);k(o,{id:"sequential",options:["InputBlock","ImageEncoderBlock"],children:(s,M)=>{var r=X(),d=J(r);U(d,{id:"sequential",option:"InputBlock",children:(n,m)=>{a(n,{code:"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",highlighted:`<span class="hljs-keyword">from</span> diffusers.modular_pipelines <span class="hljs-keyword">import</span> ModularPipelineBlocks, InputParam, OutputParam

<span class="hljs-keyword">class</span> <span class="hljs-title class_">InputBlock</span>(<span class="hljs-title class_ inherited__">ModularPipelineBlocks</span>):

<span class="hljs-meta">    @property</span>
    <span class="hljs-keyword">def</span> <span class="hljs-title function_">inputs</span>(<span class="hljs-params">self</span>):
        <span class="hljs-keyword">return</span> [
            InputParam(name=<span class="hljs-string">&quot;prompt&quot;</span>, type_hint=<span class="hljs-built_in">list</span>, description=<span class="hljs-string">&quot;list of text prompts&quot;</span>),
            InputParam(name=<span class="hljs-string">&quot;num_images_per_prompt&quot;</span>, type_hint=<span class="hljs-built_in">int</span>, description=<span class="hljs-string">&quot;number of images per prompt&quot;</span>),
        ]

<span class="hljs-meta">    @property</span>
    <span class="hljs-keyword">def</span> <span class="hljs-title function_">intermediate_outputs</span>(<span class="hljs-params">self</span>):
        <span class="hljs-keyword">return</span> [
            OutputParam(name=<span class="hljs-string">&quot;batch_size&quot;</span>, description=<span class="hljs-string">&quot;calculated batch size&quot;</span>),
        ]

<span class="hljs-meta">    @property</span>
    <span class="hljs-keyword">def</span> <span class="hljs-title function_">description</span>(<span class="hljs-params">self</span>):
        <span class="hljs-keyword">return</span> <span class="hljs-string">&quot;A block that determines batch_size based on the number of prompts and num_images_per_prompt argument.&quot;</span>

    <span class="hljs-keyword">def</span> <span class="hljs-title function_">__call__</span>(<span class="hljs-params">self, components, state</span>):
        block_state = <span class="hljs-variable language_">self</span>.get_block_state(state)
        batch_size = <span class="hljs-built_in">len</span>(block_state.prompt)
        block_state.batch_size = batch_size * block_state.num_images_per_prompt
        <span class="hljs-variable language_">self</span>.set_block_state(state, block_state)
        <span class="hljs-keyword">return</span> components, state`,lang:"py",wrap:!1})},$$slots:{default:!0}});var A=l(d,2);U(A,{id:"sequential",option:"ImageEncoderBlock",children:(n,m)=>{a(n,{code:"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",highlighted:`<span class="hljs-keyword">import</span> torch
<span class="hljs-keyword">from</span> diffusers.modular_pipelines <span class="hljs-keyword">import</span> ModularPipelineBlocks, InputParam, OutputParam

<span class="hljs-keyword">class</span> <span class="hljs-title class_">ImageEncoderBlock</span>(<span class="hljs-title class_ inherited__">ModularPipelineBlocks</span>):

<span class="hljs-meta">    @property</span>
    <span class="hljs-keyword">def</span> <span class="hljs-title function_">inputs</span>(<span class="hljs-params">self</span>):
        <span class="hljs-keyword">return</span> [
            InputParam(name=<span class="hljs-string">&quot;image&quot;</span>, type_hint=<span class="hljs-string">&quot;PIL.Image&quot;</span>, description=<span class="hljs-string">&quot;raw input image to process&quot;</span>),
            InputParam(name=<span class="hljs-string">&quot;batch_size&quot;</span>, type_hint=<span class="hljs-built_in">int</span>),
        ]

<span class="hljs-meta">    @property</span>
    <span class="hljs-keyword">def</span> <span class="hljs-title function_">intermediate_outputs</span>(<span class="hljs-params">self</span>):
        <span class="hljs-keyword">return</span> [
            OutputParam(name=<span class="hljs-string">&quot;image_latents&quot;</span>, description=<span class="hljs-string">&quot;latents representing the image&quot;</span>
        ]

<span class="hljs-meta">    @property</span>
    <span class="hljs-keyword">def</span> <span class="hljs-title function_">description</span>(<span class="hljs-params">self</span>):
        <span class="hljs-keyword">return</span> <span class="hljs-string">&quot;Encode raw image into its latent presentation&quot;</span>

    <span class="hljs-keyword">def</span> <span class="hljs-title function_">__call__</span>(<span class="hljs-params">self, components, state</span>):
        block_state = <span class="hljs-variable language_">self</span>.get_block_state(state)
        <span class="hljs-comment"># 模拟处理图像</span>
        <span class="hljs-comment"># 这将改变所有块的图像状态，从PIL图像变为张量</span>
        block_state.image = torch.randn(<span class="hljs-number">1</span>, <span class="hljs-number">3</span>, <span class="hljs-number">512</span>, <span class="hljs-number">512</span>)
        block_state.batch_size = block_state.batch_size * <span class="hljs-number">2</span>
        block_state.image_latents = torch.randn(<span class="hljs-number">1</span>, <span class="hljs-number">4</span>, <span class="hljs-number">64</span>, <span class="hljs-number">64</span>)
        <span class="hljs-variable language_">self</span>.set_block_state(state, block_state)
        <span class="hljs-keyword">return</span> components, state`,lang:"py",wrap:!1})},$$slots:{default:!0}}),e(s,r)},$$slots:{default:!0}});var i=l(o,6);a(i,{code:"ZnJvbSUyMGRpZmZ1c2Vycy5tb2R1bGFyX3BpcGVsaW5lcyUyMGltcG9ydCUyMFNlcXVlbnRpYWxQaXBlbGluZUJsb2NrcyUyQyUyMEluc2VydGFibGVEaWN0JTBBJTBBYmxvY2tzX2RpY3QlMjAlM0QlMjBJbnNlcnRhYmxlRGljdCgpJTBBYmxvY2tzX2RpY3QlNUIlMjJpbnB1dCUyMiU1RCUyMCUzRCUyMGlucHV0X2Jsb2NrJTBBYmxvY2tzX2RpY3QlNUIlMjJpbWFnZV9lbmNvZGVyJTIyJTVEJTIwJTNEJTIwaW1hZ2VfZW5jb2Rlcl9ibG9jayUwQSUwQWJsb2NrcyUyMCUzRCUyMFNlcXVlbnRpYWxQaXBlbGluZUJsb2Nrcy5mcm9tX2Jsb2Nrc19kaWN0KGJsb2Nrc19kaWN0KQ==",highlighted:`<span class="hljs-keyword">from</span> diffusers.modular_pipelines <span class="hljs-keyword">import</span> SequentialPipelineBlocks, InsertableDict

blocks_dict = InsertableDict()
blocks_dict[<span class="hljs-string">&quot;input&quot;</span>] = input_block
blocks_dict[<span class="hljs-string">&quot;image_encoder&quot;</span>] = image_encoder_block

blocks = SequentialPipelineBlocks.from_blocks_dict(blocks_dict)`,lang:"py",wrap:!1});var y=l(i,4);a(y,{code:"cHJpbnQoYmxvY2tzKSUwQXByaW50KGJsb2Nrcy5kb2Mp",highlighted:`<span class="hljs-built_in">print</span>(blocks)
<span class="hljs-built_in">print</span>(blocks.doc)`,lang:"py",wrap:!1});var h=l(y,2);G(h,{source:"https://github.com/huggingface/diffusers/blob/main/docs/source/zh/modular_diffusers/sequential_pipeline_blocks.md"}),E(2),e(T,p),C()}export{F as component};